{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Tutorial for analyzing instrumental learning data with the HDDMrl module\n", "This is a tutorial for using the HDDMrl module to simultaneously estimate reinforcement learning parameters and decision parameters within a fully hierarchical bayesian estimation framework, including steps for sampling, assessing convergence, model fit, parameter recovery, and posterior predictive checks (model validation). The module uses the reinforcement learning drift diffusion model (RLDDM), a reinforcement learning model that replaces the standard “softmax” choice function with a drift diffusion process. The softmax and drift diffusion process is equivalent for capturing choice proportions, but the DDM also takes RT distributions into account; options are provided to also only fit RL parameters without RT. The RLDDM estimates trial-by-trial drift rate as a scaled difference in expected rewards (expected reward for upper bound alternative minus expected reward for lower bound alternative). Expected rewards are updated with a delta learning rule using either a single learning rate or with separate learning rates for positive and negative prediction errors. The model also includes the standard DDM-parameters. The RLDDM is described in detail in [Pedersen, Frank & Biele (2017).](http://ski.clps.brown.edu/papers/PedersenEtAl_RLDDM.pdf) (Note this approach differs from Frank et al (2015) who used HDDM to fit instrumental learning but did not allow for simultaneous estimation of learning parameters). \n", "\n", "## OUTLINE \n", "\n", "[1. Background](#1.-Background)
\n", "[2. Installing the module](#2.-Installing-the-module)
\n", "[3. How the RLDDM works](#3.-How-the-RLDDM-works)
\n", "[4. Structuring data](#4.-Structuring-data)
\n", "[5. Running basic model](#5.-Running-basic-model)
\n", "[6. Checking results](#6.-Checking-results)
\n", "[7. Posterior predictive checks](#7.-Posterior-predictive-checks)
\n", "[8. Parameter recovery](#8.-Parameter-recovery)
\n", "[9. Separate learning rates for positive and negative prediction errors](#9.-Separate-learning-rates-for-positive-and-negative-prediction-errors)
\n", "[10. depends_on vs. split_by](#10.-depends_on-vs.-split_by)
\n", "[11. Probabilistic binary outcomes vs. normally distributed outcomes](#11.-Probabilistic-binary-outcomes-vs.-normally-distributed-outcomes)
\n", "[12. HDDMrlRegressor](#12.-HDDMrlRegressor)
\n", "[13. Regular RL without RT](#13.-Regular-RL-without-RT)\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 1. Background\n", "Traditional RL models typically assume static decision processes (e.g. softmax), and the DDM typically assumes static decision variables (stimuli are modeled with the same drift rate across trials). The RLDDM combines dynamic decision variables from RL and dynamic choice process from DDM by assuming trial-by-trial drift rate that depends on the difference in expected rewards, which are updated on each trial by a rate of the prediction error dependent on the learning rate. The potential benefit of the RLDDM is thus to gain a better insight into decision processes in instrumental learning by also accounting for speed of decision making." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 2. Installing the module\n", "The new version of HDDM (version 0.7.1) that includes the RLDDM is currently not uploaded to conda. So to install you would either have to use pip:
\n", "'pip install hddm'
\n", "OR
\n", "docker: 'pull madslupe/hddm', which runs hddm in jupyter notebook
" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 3. How the RLDDM works\n", "The main idea of the RLDDM is that reward-based choices can be captured by an accumulation-to-bound process where drift rate is proportional to the difference in expected reward between options, and where expected rewards subsequently are updated in a trial-by-trial basis via reinforcement learning.

\n", "__drift rate on each trial depends on difference in expected rewards for the two alternatives (q_up and q_low):__
\n", "drift rate = (q_up - q_low) * scaling

\n", "_where the scaling parameter describes the weight to put on the difference in q-values._

\n", "__expeceted reward (q) for chosen option is updated according to delta learning rule :__
\n", "q_chosen = q_chosen + alpha * (feedback-q_chosen)

\n", "_where alpha weights the rate of learning on each trial._

\n", "So in principle all you need is the Wiener first passage time likelihood-function. The reason why HDDM is useful (and hence also HDDMrl) is that it automates the process of setting up and running your model, which tends to be very time consuming. So after structuring the data it is simple to run a model with HDDMrl. In particular it separates subjects and conditions (using the split_by-column, see next section) so that the updating process works correctly, which can be especially difficult to do for RL models. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 4. Structuring data\n", "The HDDMrl module was created to make it easier to model instrumental learning data with the RLDDM. If you are familiar with using HDDM it shouldn't be a big step to start using HDDMrl. Please refresh your memory by starting with the [tutorial for HDDM](http://ski.clps.brown.edu/hddm_docs/index.html) first (especially critical if you have not used HDDM at all). Running HDDMrl does require a few extra steps compared to HDDM, and because the model includes increased potential for parameter colinearity (typically learning rate and the scaling parameter on drift rate) it is even more important to assess model fit, which will be covered below. Here are the most important steps for structuring your dataframe:\n", "1. Sort trials in ascending order within subject and condition, to ensure proper updating of expected rewards.\n", "2. Include a column called __'split_by'__ which identifies the different task conditions (__as integers__), to ensure reward updating will work properly for each condition without mixing values learned from one trial type to another. \n", "3. Code the response column with [stimulus-coding] (http://ski.clps.brown.edu/hddm_docs/howto.html#code-subject-responses). Although stimulus-coding and accuracy-coding often are the same in instrumental learning it is important that the upper and lower boundaries are represented by the same alternative within a condition, because expected rewards are linked to the thresholds/boundaries.\n", "4. __feedback__-column. This should be the reward received for the chosen option on each trial.\n", "5. __q_init__. Adjusting initial q-values is something that can improve model fit quite a bit. To allow the user to set their own initial values we therefore require that the dataframe includes a column called q_init. The function will set all initial q-values according to the first value in q_init. So this is not the most elegant method of allowing users to set inital value for expected rewards, but it works for now.\n", "\n", "#### Required columns in data:\n", "* __rt__: in seconds, same as in HDDM\n", "* __response__: 0 or 1. defines chosen stimulus, not accuracy.\n", "* __split_by__: needs to be an integer. Split_by defines conditions (trial-types) that should have the same variables (e.g. Q values) within subject: the trials need to be split by condition to ensure proper updating of expected rewards that do not bleed over into other trial-types. (e.g. if you have stimulus A and get reward you want that updated value to impact choice only for the next stimulus A trial but not necessarily the immediate trial afterwards, which may be of a different condition) \n", "* __subj_idx__: same as in HDDM, but even more important here because it is used to split trials\n", "* __feedback__: feedback on the current trial. can be any value.\n", "* __q_init__: used to initialize expected rewards. can be any value, but an unbiased initial value should be somewhere between the minimum and and maximum reward values (e.g. 0.5 for tasks with rewards of 0 and 1)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 5. Running basic model\n", "To illustrate how to run the model we will use example data from the learning phase of the probabilistic selection task (PST). During the learning phase of the PST subjects choose between two stimuli presented as Hiragana-letters (here represented as letters from the latin alphabet). There are three conditions with different probabilities of receiving reward (feedback=1) and non-reward (feedback=0). In the AB condition A is rewarded with 80% probability, B with 20%. In the CD condition C is rewarded with 70% probability and D with 30%, while in the EF condition E is rewarded with a 60% probability and F with 40%. The dataset is included in the data-folder in your installation of HDDM." ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "/home/krishn/anaconda3/envs/hddm/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:526: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n", " _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n", "/home/krishn/anaconda3/envs/hddm/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:527: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n", " _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n", "/home/krishn/anaconda3/envs/hddm/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:528: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n", " _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n", "/home/krishn/anaconda3/envs/hddm/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:529: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n", " _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n", "/home/krishn/anaconda3/envs/hddm/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:530: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n", " _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n", "/home/krishn/anaconda3/envs/hddm/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:535: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n", " np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "tanh\n", "tanh\n", "tanh\n", "linear\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "/home/krishn/anaconda3/envs/hddm/lib/python3.6/site-packages/IPython/parallel.py:13: ShimWarning: The `IPython.parallel` package has been deprecated since IPython 4.0. You should import from ipyparallel instead.\n", " \"You should import from ipyparallel instead.\", ShimWarning)\n" ] } ], "source": [ "# import\n", "import pandas as pd\n", "import numpy as np\n", "import hddm\n", "from scipy import stats\n", "import seaborn as sns\n", "import matplotlib.pyplot as plt\n", "import pymc\n", "import kabuki\n", "\n", "sns.set(style=\"white\")\n", "%matplotlib inline\n", "from tqdm import tqdm\n", "import warnings\n", "\n", "warnings.filterwarnings(\"ignore\", category=np.VisibleDeprecationWarning)" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
subj_idxresponsecondrttrialsplit_byfeedbackq_init
0420.0CD1.2551.010.00.5
1421.0EF0.7781.020.00.5
2421.0AB0.6471.001.00.5
3421.0AB0.7502.001.00.5
4420.0EF0.7722.021.00.5
\n", "
" ], "text/plain": [ " subj_idx response cond rt trial split_by feedback q_init\n", "0 42 0.0 CD 1.255 1.0 1 0.0 0.5\n", "1 42 1.0 EF 0.778 1.0 2 0.0 0.5\n", "2 42 1.0 AB 0.647 1.0 0 1.0 0.5\n", "3 42 1.0 AB 0.750 2.0 0 1.0 0.5\n", "4 42 0.0 EF 0.772 2.0 2 1.0 0.5" ] }, "execution_count": 2, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# load data. you will find this dataset in your hddm-folder under hddm/examples/rlddm_data.csv\n", "data = hddm.load_csv(\"rlddm_data.csv\")\n", "# check structure\n", "data.head()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The columns in the datafile represent: __subj_idx__ (subject id), __response__ (1=best option), __cond__ (identifies condition, but not used in model), __rt__ (in seconds), 0=worst option), __trial__ (the trial-iteration for a subject within each condition), __split_by__ (identifying condition, used for running the model), __feedback__ (whether the response given was rewarded or not), __q_init__ (the initial q-value used for the model, explained above)." ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "{}\n", "()\n", "No model attribute --> setting up standard HDDM\n", "Includes supplied: ()\n", "printing self.nn\n", "False\n", "Set model to ddm\n", " [-----------------100%-----------------] 1500 of 1500 complete in 143.7 sec mean std 2.5q 25q 50q 75q 97.5q mc err\n", "a 1.7433 0.0813138 1.58494 1.68967 1.74494 1.7954 1.9187 0.00292143\n", "a_std 0.406522 0.0610479 0.312902 0.3623 0.401829 0.441929 0.558704 0.00266828\n", "a_subj.3 2.00633 0.10129 1.83542 1.93472 1.99867 2.06634 2.23372 0.00443095\n", "a_subj.4 1.89855 0.0539594 1.79832 1.86133 1.89825 1.93209 2.01204 0.002394\n", "a_subj.5 1.48664 0.0741329 1.34377 1.4389 1.48745 1.53229 1.6436 0.00351514\n", "a_subj.6 2.39376 0.100291 2.19236 2.32632 2.39406 2.4608 2.5925 0.00447612\n", "a_subj.8 2.49353 0.132293 2.23753 2.40396 2.49135 2.57203 2.77695 0.00695635\n", "a_subj.12 1.90207 0.0684117 1.77964 1.85209 1.89715 1.94514 2.05585 0.00313201\n", "a_subj.17 1.40287 0.0822611 1.22774 1.35337 1.40235 1.45487 1.56221 0.00487811\n", "a_subj.18 1.90812 0.125013 1.67796 1.81644 1.90952 1.99562 2.15789 0.00679012\n", "a_subj.19 2.08064 0.085651 1.92067 2.02268 2.08202 2.13764 2.24183 0.00446387\n", "a_subj.20 1.71228 0.0794146 1.55761 1.65782 1.70888 1.76279 1.8681 0.00369196\n", "a_subj.22 1.19307 0.0335414 1.12915 1.17129 1.19207 1.21472 1.26232 0.00158518\n", "a_subj.23 1.93739 0.0895423 1.76888 1.87749 1.93578 1.99651 2.12028 0.00376337\n", "a_subj.24 1.68377 0.0952974 1.514 1.62176 1.68111 1.74522 1.87941 0.00482898\n", "a_subj.26 2.23807 0.0803151 2.08342 2.18239 2.23925 2.29066 2.40204 0.00362931\n", "a_subj.33 1.56624 0.066382 1.44514 1.51945 1.56395 1.61366 1.70245 0.00274963\n", "a_subj.34 1.81803 0.0883775 1.64683 1.75351 1.82055 1.87853 1.99149 0.00376558\n", "a_subj.35 1.83533 0.0975695 1.64512 1.77156 1.83433 1.9008 2.03218 0.00488217\n", "a_subj.36 1.31686 0.0685052 1.19867 1.2683 1.31396 1.35797 1.46606 0.00292263\n", "a_subj.39 1.57041 0.0530064 1.47084 1.53222 1.57151 1.60465 1.68414 0.00183899\n", "a_subj.42 1.72821 0.0757661 1.58521 1.67929 1.7261 1.77948 1.8817 0.00388646\n", "a_subj.50 1.46984 0.0556356 1.3657 1.42899 1.46951 1.51015 1.57257 0.00366184\n", "a_subj.52 2.1227 0.0885946 1.95207 2.06009 2.12394 2.18527 2.29723 0.00371883\n", "a_subj.56 1.47025 0.0482088 1.37467 1.44058 1.46858 1.50214 1.56976 0.00245215\n", "a_subj.59 1.31005 0.0817522 1.16181 1.25224 1.30721 1.36636 1.46954 0.00524295\n", "a_subj.63 1.8639 0.0883703 1.71047 1.801 1.8624 1.92234 2.04382 0.00375231\n", "a_subj.71 1.24626 0.0340768 1.17937 1.22192 1.24647 1.27036 1.31251 0.00140437\n", "a_subj.75 0.97873 0.0365919 0.913123 0.951568 0.977399 1.00419 1.05616 0.00179393\n", "a_subj.80 2.187 0.0965787 2.00696 2.12386 2.18203 2.24945 2.37954 0.00388193\n", "v 4.08209 0.737245 2.64887 3.57663 4.09291 4.5691 5.5414 0.0458143\n", "v_std 3.05484 0.575751 2.07762 2.64448 2.98722 3.40931 4.34097 0.0386874\n", "v_subj.3 5.3036 2.4037 1.33138 3.46797 5.08761 6.83865 10.67 0.180208\n", "v_subj.4 0.936614 0.141934 0.679051 0.839064 0.931719 1.02271 1.23745 0.00588382\n", "v_subj.5 5.06788 2.26694 1.35156 3.39827 4.84072 6.55001 10.3183 0.158873\n", "v_subj.6 4.58494 2.4764 1.19679 2.43696 4.25878 6.21347 9.97977 0.218163\n", "v_subj.8 3.1593 1.1972 1.73287 2.40894 2.89231 3.50444 6.66669 0.0985572\n", "v_subj.12 1.83104 0.224408 1.44052 1.67555 1.81061 1.9754 2.30106 0.0107878\n", "v_subj.17 6.93416 2.22622 3.44608 5.39757 6.56186 8.19763 12.2283 0.168861\n", "v_subj.18 6.12272 1.59534 3.55255 4.97474 5.93287 7.09307 9.95941 0.122813\n", "v_subj.19 11.7997 1.9933 8.60538 10.4711 11.5523 12.8174 16.4139 0.156147\n", "v_subj.20 4.72675 1.23807 2.99018 3.87997 4.53657 5.3181 7.96896 0.104607\n", "v_subj.22 2.19815 0.283286 1.68683 1.9937 2.18769 2.39485 2.73678 0.0128246\n", "v_subj.23 2.4331 1.50181 0.959062 1.4395 1.90909 2.89597 6.60491 0.128198\n", "v_subj.24 5.12042 0.753118 3.67884 4.60547 5.10993 5.62647 6.63844 0.0334464\n", "v_subj.26 0.602543 0.863859 0.24547 0.404374 0.466825 0.555059 1.98844 0.0705156\n", "v_subj.33 2.73658 2.10267 0.282828 1.14428 2.1263 3.89181 8.07948 0.120839\n", "v_subj.34 4.11408 2.32502 1.11361 2.29809 3.65292 5.45819 9.77123 0.15346\n", "v_subj.35 3.14171 1.87384 1.09744 1.80252 2.57169 3.96941 7.90195 0.155472\n", "v_subj.36 3.87701 2.00542 1.52318 2.39526 3.27531 4.79694 9.1923 0.161708\n", "v_subj.39 2.0298 1.62953 0.34513 0.935087 1.45793 2.58893 6.68562 0.107378\n", "v_subj.42 5.16806 1.89934 2.28067 3.90109 4.91005 6.18106 9.81468 0.164759\n", "v_subj.50 5.57649 1.79009 2.85308 4.24585 5.34125 6.59774 9.84759 0.141939\n", "v_subj.52 3.18577 1.84878 1.16424 1.85057 2.71706 3.88391 8.41286 0.159\n", "v_subj.56 2.05017 1.91277 0.258263 0.703241 1.36834 2.86095 7.32663 0.14203\n", "v_subj.59 10.9111 1.82836 7.62548 9.62643 10.8841 12.1672 14.6141 0.133133\n", "v_subj.63 2.22187 1.32504 0.988662 1.4741 1.91208 2.45281 6.4643 0.105276\n", "v_subj.71 2.17006 1.78241 0.603637 1.08317 1.48305 2.50471 7.27104 0.149917\n", "v_subj.75 3.71969 0.552382 2.70805 3.34183 3.69385 4.06496 4.9227 0.0277269\n", "v_subj.80 4.03709 2.17552 0.986179 2.27813 3.66944 5.33457 9.111 0.170157\n", "t 0.432885 0.0415403 0.356521 0.404121 0.431653 0.462548 0.511254 0.00178714\n", "t_std 0.232643 0.0394899 0.169301 0.203101 0.227423 0.257429 0.315877 0.00179352\n", "t_subj.3 1.06317 0.0271769 0.999802 1.04743 1.06651 1.08276 1.10684 0.00129274\n", "t_subj.4 0.527059 0.0179398 0.488522 0.515254 0.528076 0.539991 0.557209 0.000826279\n", "t_subj.5 0.528425 0.0158093 0.493437 0.519202 0.529042 0.538151 0.559482 0.000756408\n", "t_subj.6 0.398728 0.0316425 0.330672 0.379382 0.399754 0.41924 0.461281 0.00145695\n", "t_subj.8 0.663496 0.0406703 0.573784 0.638121 0.667493 0.692995 0.734004 0.00220312\n", "t_subj.12 0.408965 0.0147454 0.378986 0.398904 0.410359 0.41963 0.436432 0.000711611\n", "t_subj.17 0.500941 0.0182598 0.472563 0.491633 0.499139 0.505433 0.563134 0.00142979\n", "t_subj.18 0.433627 0.0282406 0.371316 0.416532 0.438131 0.455078 0.47849 0.00153929\n", "t_subj.19 0.419058 0.0150433 0.387912 0.409382 0.419564 0.429151 0.4471 0.000741052\n", "t_subj.20 0.51628 0.0127393 0.491093 0.507661 0.516726 0.525528 0.538053 0.000610317\n", "t_subj.22 0.33496 0.00518669 0.324114 0.331928 0.335304 0.338463 0.344235 0.000253281\n", "t_subj.23 0.481226 0.0226842 0.431716 0.466665 0.483849 0.497623 0.519431 0.000941342\n", "t_subj.24 0.453762 0.013039 0.425115 0.445645 0.454502 0.462701 0.47683 0.000579678\n", "t_subj.26 0.440285 0.0338985 0.372521 0.414917 0.443772 0.468245 0.49323 0.00164635\n", "t_subj.33 0.194727 0.0156474 0.164383 0.184158 0.194446 0.204649 0.225083 0.000671091\n", "t_subj.34 0.344477 0.0311183 0.276129 0.327923 0.345049 0.361748 0.413056 0.00126463\n", "t_subj.35 0.326443 0.0258584 0.273692 0.309526 0.32619 0.343838 0.377765 0.00119399\n", "t_subj.36 0.44915 0.0134952 0.415288 0.443814 0.452192 0.458502 0.466911 0.000650595\n", "t_subj.39 0.619389 0.0110796 0.595795 0.612534 0.620065 0.627238 0.638565 0.000401916\n", "t_subj.42 0.393542 0.0155789 0.362458 0.383731 0.394086 0.404417 0.422266 0.000830944\n", "t_subj.50 0.518358 0.0214237 0.479266 0.498422 0.524998 0.536913 0.548587 0.00150307\n", "t_subj.52 0.514745 0.0238354 0.46713 0.499788 0.515337 0.530868 0.56368 0.00104662\n", "t_subj.56 0.118365 0.0108243 0.0959725 0.11129 0.119001 0.12587 0.139238 0.00054126\n", "t_subj.59 0.37923 0.0228073 0.332184 0.363319 0.377838 0.39906 0.415741 0.00167972\n", "t_subj.63 0.478244 0.0193531 0.437445 0.466752 0.479564 0.491179 0.513794 0.000800474\n", "t_subj.71 0.162673 0.0054732 0.150777 0.159311 0.162957 0.166633 0.172557 0.000235596\n", "t_subj.75 0.260777 0.00626259 0.244606 0.258207 0.262153 0.264765 0.269822 0.000295097\n", "t_subj.80 0.0417991 0.0176072 0.0114696 0.028693 0.041848 0.0536039 0.0781576 0.000700406\n", "alpha -2.46715 0.356073 -3.157 -2.69308 -2.48242 -2.24392 -1.74121 0.0195181\n", "alpha_std 1.46992 0.275563 1.0209 1.28032 1.45203 1.61981 2.09465 0.0155966\n", "alpha_subj.3 -3.43548 0.740258 -4.48635 -3.90149 -3.56324 -3.08342 -1.45227 0.0608327\n", "alpha_subj.4 0.191786 0.595174 -0.801725 -0.213263 0.126076 0.52319 1.65872 0.0268693\n", "alpha_subj.5 -3.28297 0.771161 -4.52565 -3.7898 -3.39336 -2.93093 -1.27709 0.0562957\n", "alpha_subj.6 -3.69185 0.841373 -4.92836 -4.33939 -3.87057 -3.1469 -1.67019 0.0743001\n", "alpha_subj.8 -2.29198 0.611898 -3.76992 -2.61173 -2.22875 -1.86592 -1.29704 0.047517\n", "alpha_subj.12 -0.780933 0.425905 -1.62084 -1.05694 -0.786673 -0.502726 0.0835042 0.0206346\n", "alpha_subj.17 -2.81596 0.484016 -3.76465 -3.14894 -2.81048 -2.5293 -1.77409 0.0371539\n", "alpha_subj.18 -2.75671 0.392851 -3.44184 -3.03159 -2.77331 -2.50711 -1.88007 0.030011\n", "alpha_subj.19 -3.7218 0.226315 -4.21729 -3.85334 -3.70381 -3.58146 -3.29365 0.0171907\n", "alpha_subj.20 -2.37897 0.438575 -3.25441 -2.65439 -2.3874 -2.09008 -1.51432 0.0362439\n", "alpha_subj.22 -0.975383 0.212324 -1.36486 -1.12002 -0.979027 -0.845113 -0.547454 0.00859101\n", "alpha_subj.23 -2.29157 0.971566 -4.14016 -2.95807 -2.24846 -1.50684 -0.582361 0.0809592\n", "alpha_subj.24 -1.49115 0.231074 -1.95491 -1.64755 -1.47849 -1.33537 -1.05145 0.00874484\n", "alpha_subj.26 0.477848 1.54335 -5.25248 0.166103 0.700563 1.2395 2.52254 0.12061\n", "alpha_subj.33 -3.25956 1.33119 -5.54438 -4.22104 -3.29747 -2.31559 -0.650837 0.0832202\n", "alpha_subj.34 -2.9963 0.924299 -4.55411 -3.65087 -3.1139 -2.40745 -1.05377 0.0637804\n", "alpha_subj.35 -2.68266 0.985176 -4.42859 -3.41642 -2.76104 -1.94719 -0.782148 0.0807771\n", "alpha_subj.36 -2.35113 0.964568 -3.91305 -3.07013 -2.41253 -1.6805 -0.415103 0.0775148\n", "alpha_subj.39 -3.07825 1.23011 -5.28098 -4.03919 -3.06089 -2.16057 -0.719406 0.0801558\n", "alpha_subj.42 -3.14098 0.59356 -4.12935 -3.52327 -3.22374 -2.80894 -1.82638 0.0516964\n", "alpha_subj.50 -3.78284 0.656364 -4.82933 -4.26283 -3.87635 -3.39796 -2.22331 0.0510745\n", "alpha_subj.52 -3.24081 0.75318 -4.72941 -3.78718 -3.26367 -2.68309 -1.89825 0.0642394\n", "alpha_subj.56 -3.88438 1.4086 -6.07973 -4.95683 -4.04305 -2.87882 -0.949276 0.107409\n", "alpha_subj.59 -2.54493 0.225305 -2.98248 -2.69149 -2.55072 -2.39734 -2.06605 0.0149846\n", "alpha_subj.63 -2.04163 0.949775 -4.16 -2.56008 -1.95309 -1.44029 -0.246927 0.0683846\n", "alpha_subj.71 -2.84608 1.54153 -5.48198 -4.07678 -2.63727 -1.8612 0.43124 0.123145\n", "alpha_subj.75 -1.37871 0.44339 -2.24472 -1.67589 -1.36566 -1.06067 -0.558518 0.0223029\n", "alpha_subj.80 -3.22538 0.795376 -4.50105 -3.82306 -3.32231 -2.75346 -1.47653 0.0631666\n", "DIC: 10127.947410\n", "deviance: 10059.166818\n", "pD: 68.780592\n" ] } ], "source": [ "# run the model by calling hddm.HDDMrl (instead of hddm.HDDM for normal HDDM)\n", "m = hddm.HDDMrl(data)\n", "# set sample and burn-in\n", "m.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")\n", "# print stats to get an overview of posterior distribution of estimated parameters\n", "m.print_stats()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Interpreting output from print_stats:__
\n", "The model estimates group mean and standard deviation parameters and subject parameters for the following latent variables:
\n", "a = decision threshold
\n", "v = scaling parameter
\n", "t = non-decision time
\n", "alpha = learning rate, note that it's not bound between 0 and 1. to transform take inverse logit: np.exp(alpha)/(1+np.exp(alpha))
\n", "The columns represent the mean, standard deviation and quantiles of the approximated posterior distribution of each parameter" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### HDDMrl vs. HDDM\n", "__There are a few things to note that is different from the normal HDDM model.__
\n", "First of all, the estimated learning rate does not necessarily fall between 0 and 1. This is because it is estimated as a normal distribution for purposes of sampling hierarchically and then transformed by an inverse logit function to 0\n", "Second, the v-parameter in the output is the scaling factor that is multiplied by the difference in q-values, so it is not the actual drift rate (or rather, it is the equivalent drift rate when the difference in Q values is exactly 1)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 6. Checking results" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Plotting a\n", "Plotting a_std\n", "Plotting v\n", "Plotting v_std\n", "Plotting t\n", "Plotting t_std\n", "Plotting alpha\n", "Plotting alpha_std\n" ] }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" }, { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlwAAAFuCAYAAABURsaLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAgAElEQVR4nOzdeXwU5f3A8c/m4lbkToAKCjwi4kFUPIpQrYgtiNpWoUq1taCt1qq0VlsrP6q0Wg1Wq1ZBrdoqeBQVqDcW1BYv8ALxQRSESAC5w02S/f2xM5vZ3Znd2dnZzO7m+369WpLZOb67T8x88zzfeZ5QOBxGCCGEEEJkT1HQAQghhBBCFDpJuIQQQgghskwSLiGEEEKILJOESwghhBAiyyThEkIIIYTIMkm4hBBCCCGyTBIuIYQQQogsk4RLCCGEECLLJOESQgghhMgySbiEEEIIIbJMEi4hhBBCiCwrCToAJ0qpFsBxQA1QH3A4QoimUQyUA+9qrfcGHYxX8vtLiGYp6e+vnE24iPyyeiPoIIQQgRgCvBl0EBmQ319CNF+2v79yOeGqAXjsscfo1q1b0LEIIZrAunXruOCCC8D47z+Pye8v4M477wTgl7/8ZcCRZK53794ArFy5MuBImkYhtV1TSfX7K5cTrnqAbt260aNHj6BjEUI0rXwfhpPfX8CgQYMACuIzqKurAwrjvbhRSG0XANvfX7mccAkhhMhjY8eODToE4ZG0nf/kKUUhhBBCiCyThEsIIYQQIssyGlJUSo0A7iTyKOQDWutbbPYZBvwFKAU2aq2HZnLNXPbEK5r1m3dx5fnHBB2KEEIEbvLkyQBMmjQp4EhEuqTt/Oe5h0spVQzcA5wJHA6MVUodHrdPe+Be4Cyt9QDgBxnEmvP++eKnvPLO6qDDEEIIIUSOyaSH63hghdb6CwCl1ExgNPCJZZ8fArO01qsBtNYbMrieEEKIPCK9I/lL2s5/mSRc3YE1lu+rgcFx+/QDSpVS84F2wJ1a60czuKYQQgghRN7JpGg+ZLMtHPd9CVAJfBc4A/i9UqpfBtcUInB19Q3s2VsXdBgZWbO+llETn+O9ZeuDDkUIIZqFTHq4qoGelu97AGtt9tmotd4J7FRKvQ4cBSzP4LpCBGry9Lf44LOvmVM1OuhQPNNfbgbgjQ++4tj+XQOOpukopVoCrwMtiPz+e1prPUkp1QF4AugFrALO01pvMY65HriEyGSGV2qtXwog9Lw0Y8YMQOZ0ykfSdv7LpIfrXaCvUqq3UqoMGAPMjtvnOWCIUqpEKdWayJDjsgyuKUTgPvjs66BD8IFdB3WzsBc4VWt9FHA0MEIpdQJwHTBPa90XmGd8j/Eg0BhgADACuNd4YEi4sHz5cpYvl7+v85G0nf8893BpreuUUlcALxGZFuIhrfVSpdRlxuv3aa2XKaVeBD4CGohMHbHEj8CFECJdWuswsMP4ttT4X5jIAz/DjO2PAPOB3xjbZ2qt9wIrlVIriDwwtLDpos5fY8aMCToE4ZG0nf8ymodLa/088Hzctvvivr8NuC2T6wgh/BUyOrjC4fiyy8Jn9FAtAvoA92it31ZKddVa1wBorWuUUl2M3bsDb1kOrza2CReUUkGHIDyStvNf3sw0/+qrr7Jp06aE7bNmzWLfvn0BRCRE/go12xFF0FrXa62PJlJ3erxS6ogku7t5OEgIIVLK+4TrmWeeYf/+/QnbGxoamiIsIfJac84ctNZbiQwdjgDWK6XKAYx/zTkD3TwcJBwsWrSIRYsWBR2GrX376/Py3E0ll9suX2U0pNhU1qxZwxtvvMFnn33GSSedxMSJEwF4//33WbZsGePHj+eMM85g+/btVFdXs2nTJn79618zffp01q9fT11dHVVVVVRUVLBgwQLuvfdeSktLOf/88xk5ciQ333wzy5cvp6SkhD/96U9069Yt4HcsRLaZY4rBRtHUlFKdgf1a661KqVbAt4FbiTzwcxFwi/Hvc8Yhs4HHlVJTgQqgL/COnzHNmjWLk08+ma5dE58WHTduHNdeey0DBw6M2f7xxx/z3HPPccMNN9ies7q6mvfff59Ro0b5GWra5s6dC0BlZWWgcdgpKy1m1MTnUu8Yx80x+fwEsymX2y5f5UXC1bNnT4YMGcJPfvIT+vVrnMbrmGOOoX///tx33320adOGv/71r5SXl3PrrbcCcNNNN9GqVSvmzZvHE088wS9/+UuqqqqYMWMGbdq0oaGhgfnz53PAAQfwj3/8gyVLljBt2jRuvPHGoN6qEE2iGQ8plgOPGHVcRcCTWuu5SqmFwJNKqUuA1RjLkBkPAj1JZAWNOuByrbWv3RfPPPMMffv2tU24nAwcODAhCbP66quvmDt3buAJ16BBgwK9vvBO2s5/eZFwpcP8JVRfX8/tt9/OsmXL2LdvH3379mXz5s2Ul5fTpk0bAIqKilixYgWvvvoq7733HuFwWHq3RLPS3GrmtdYfAQmry2utNwGnORwzBZiSjXhefPFFlixZwq9+9StatmzJE088QcuWLRP2mTx5MrW1tUyZMoVjjz2Wt99+m4ceeoj777+fd955hylTIuGFQiH++c9/UlVVxeeff87o0aM555xzuPjii7MRfkpBJ3zCO2k7/+VNwlVSUmJblxW/vagoUpa2bNkyNm7cyOOPP868efN45ZVX6NChA+vWrWPXrl20bt2ahoYGDjnkEEaMGMHll18OYFsPJkShMTu4ws1tTDHHjBgxgscee8x22NBUX1/P008/zYIFC7j77rt5+OGHY15/6KGHuPHGG6msrGTnzp20aNGCiRMnRhMyIURuyJui+VNOOYUpU6Ywbdq0mO2nnnoqV111FU899VTM9kMOOYSvv/6aH//4xyxcGJkyp6ioiKuvvpqLLrqIcePG8e9//5tTTz2VrVu3Mm7cOMaNG8ezzz7bZO9JiMA04zHFfHP66acDMGDAAL766quE1wcNGsQtt9zCo48+Sm1tLSUlufN3dG1tLbW1tUGHITyQtvNf7vyXmcLw4cMZPnx4wnYzUYrXunVrHn/88YTtw4YNY9iwYTHbfve73/kWpxB5RTq4cl5ZWRkQ+YOxvj6xfGzChAkMHTqUBQsWcN555/H3v/+9qUN0NHXqVAAmTZoUcCQiXdJ2/subhEuIQrVYb+AbXdvRqX2rJrum9G/ljjZt2rBz507Px69evRqlFEopPvjgA1auXEm3bt0yOqdf2rZtG3QIwiNpO/9JwiVEwCZNW8gBbcp47A9nNtk1Q81zVoicdM455zBp0iTHovlUHnnkEd5++22Kioro06cPp5xyCqFQiOLiYs466yzOPffcwIrmzSl8RP6RtvOfJFxC5IDtO5t2tYSQ0cfVHJf2yTVnnHEGZ5xxhu1r//jHP6Jfd+jQgddeew2AwYMHM3jwYAB+//vf2x77yCOP+BypECITeVM0L0SukWRFCCGEW5JwCeFRQwHkWwXwFkQOmzZtWsKT5SI/SNv5T4YUhfAqHCZvy88bJ+ISImtqamqCDkF4JG3nP0m4smDWf1Zw7rf6BB2GyLKGMBQHHYRHMg2XaArjx48POgThkbSd/2RIMQv+Pndp0CE0C5MfeIv7n/kowAjyv3tIZpoX2VRRUUFFRUXQYQgPpO38JwmXyFvvLVvP3DdXBnb9fK7hCuXrUKgQQuQpSbiE8KgQnlIsgLcgctj8+fOZP39+0GEID6Tt/CcJlxAe5XWyIh1cogksWLCABQsWBB2G8EDazn9SNC8K2ubte5jzxhf84LS+tG5Z6uu587mHS/It0RSGDh0adAjCI2k7/0nCJQraa++t4enXPuPAti04e+ihvp7bj3wr6KQt6OuLwjZs2LCgQxAeSdv5T4YURUH7essuAIqL/O/T8SNVCSrfia6lKPmWEEI0CUm4REHbsWs/AMXFWUi4fMhWgst3ZFBRZN/atWtZu3Zt0GEID6Tt/CcJlyho9cbcDfX16aU2Cz+u4dq/vpE0qfKldyiPu5iuv/dNqh5bFHQYIodNnz6d6dOnBx2G8EDazn9SwyUKmjl0Vp/mpFm3P7aIffvr2bmnjrat7Ivt87mHy48hxSWfbwJg4gWVPkQkClF5eXnQIQiPpO38JwlXltz04Nv8/pLBQYeRt5av3kKrFiX07Nouo/OEjMyioaEhrePaty1jw5bdbNm+J0nClVFovp3DCxlQFE1hwoQJQYcgPJK2858MKWbJO5+sCzqElO5+6gOemrc86DBsTbzzdX7+59cyPo+ZWKTbw2VOIWHWgNnx5wm/gJ9SlKV9hBCiSUgPVzP20ltfAvCD0/oFHEn2mD1c6SZcpoZkNVyezhh3jsCeUpQ+LiGEaErSw5VnZrz0KctXb3G177799Z4TjUKTbtF8skTLlM81XMi0EKIJVFVVUVVVFXQYwgNpO/8164Rr15797N5bF3QYaXn8Zc3EO193te/3rpvL/01fmOWIcpuZONWnWcNlJiJJe7gySFbq6ht4YeEqGgJKiL30b22p3cO+/fUALP1ik78BiYK0Y8cOduzYEXQYwgNpO/8V1JDi9p37aN2yhJJid3nk+b97HoA5VaOzGVagPlj+ddAhBMpMmNJNbKK9V0kOc9ML5mTWf1bwjxeWsb+u3vM5XntvDWs37uDCEf09nyMdP/q/lziqbyd+cFo/brjvf01yTZHfrrnmmqBDEB5J2/mvoHq4LrjxBaY+vjjoMFJavW47L/xvJS8sXMWuPc5F2U1t07bdjLnheb5ct93T8Q/PXconKzPr+di4dTdr1tdmdA4rM9FKNbT65KvL+ePD70S/d9PDlcl44JbaPQDs2es94bpjxmKeeCWzhx7SHRb98LONbNq2O6NriuajXbt2tGuX2ZPGIhjSdv4rmITLvHG88cFXvp63vr6BPT4PO/6iaj73/usj7n36Q87/3fPMX7TG1/N79daSdezcvZ9/v7nS0/H/+s8KfnP3mxnF8OObXvbl6UST24TrHy8sY+HHNdGfo2z3cNUZNWUlWZgB3w2zaN7LO5CCeyGESF9GQ4pKqRHAnUAx8IDW+pa414cBzwHmHXyW1voPmVzTifXe19AQpijF2nlua1DumPE+C96v9nXYMX54q+rxxQyr7Bmzbc++OlqWxTZPthcaNj8zL4lELiyCXN8QTlgz0Qyrvt5dDdfO3ftp27osmohkklQlY8ZTVBTs3zyLP91AfX0DxS6H4cG5/iscDtMQzs66lSI/zZkzB4BRo0YFHIlIl7Sd/zz/tldKFQP3AGcChwNjlVKH2+z6htb6aON/WUm2IPaG/96n61Puf9097npiFrxfnXD+bPv0y8384Pp/896y2PeR7RCKopOEpnehpV9sYs6bX2QjJNdee28NZ/96Nus374rZ3lg07+49/eHBtwEsPV3O+3pNxmb9ZwWvvLMaCK6Hy+rFhatS7mP9mXDq4Zr9xhec/evZbN+5z6fIRL5bvHgxixfnfpmHSCRt579M/rw+Hlihtf5Ca70PmAkEVn1uvZ9m4y/sppxe4UOj0D2+HirbEZgfW7qJxHX3vMn0Z5dEv//fR02/4Kk5lLw6rv6sIc21FJet2hw5Ljqi6P9TitYEJ/5n9ZIprzDrPyuAyJOM3/vNHOa9u9rbhVza5WLI3Poz4TSiaMa5Ycsu+x1EszNy5EhGjhwZdBjCA2k7/2WScHUHrMVH1ca2eCcqpT5USr2glBqQwfWSsvZAtWkZuxTLzFc0j734aUbnr6tLb1oBq3c+Wccrb3/pev9deyI3wNZx7yPbXVzRIcUMk8s/PfIun3652Y+QXDOTgPjYvUwLEQ6Ho591so/c7TBlPGvyEj+kuGHzLv4+dykQ+TnYV9fAg7OXELSYHi6HQcXSksh7yeS/FVFYKisrqayUtTbzkbSd/zJJuOx+68bfnhYDB2utjwL+CjybwfWSst7E4nujHnvxU2a+ojM6fyZL9dz04Nvc9eQHrvffaTy52KZlXA2X5wjcaUy4Mj/Xrt2NvSY7du/nnqc/ZM++7M15Fh0OjcuQ3BbNW4XDlh4u43zLV29hwp9ejXmq1Guvp/WoZL2xfiXAtjGkSN7D4TDbduyNfm99ryGH3xqlJcUA7JeESwghEmSScFUD1krvHkDMWJLWervWeofx9fNAqVKqUwbXdBRTNJ+FnqDb/rnI93NaWete9u6LTBXQoqw4Zh83b2vdpp3s2J041cSdM99n1MTnmPCnVx2PNZMWP+rVrMNOT766nBcXruLFhe57+dLVWPAfuz3dGi6w7+F64Lkl1GzcyRdfbYvuV+exh8vakMVJarjMV7Lx85zqjC+//SUXTnqRL2siQ7RuerjMerT9Xj8XUXC01mid2R+7IhjSdv7LJOF6F+irlOqtlCoDxgCzrTsopboppULG18cb18vKFNXWJCEXnpiD9GpZ/mutewolfGFofF8//sNLCefYtmMv4//4Kr+4LXFahVeN+pqajTsdYzATrnqfEy7zSy+TfC7WG1zt55QsRufTSifhIrGHa2ttpLenbeuyaHJXXx9m9brt/PfD9GrWrKFYWzg+RvM7u/xly/Y9aV3Tre0797FnXx0ffrYRgA9XROoJrUmf02dp9nB5TkRFwZk5cyYzZ84MOgzhgbSd/zwnXFrrOuAK4CVgGfCk1nqpUuoypdRlxm7fB5YopT4E7gLGaK2zkg1Z7wFBLZcS796nP4zpEUmmxEWhv/VtbdyWeMP9qzFsuXHbnpj1Fhd+7C4hsA5hrdu0k7r6Bnbs8vbEmbUXxOzF8TIEN2mau6WJzAQvHHevdyqar69viH5Gdkmauc0M2VwCKhwOR9uqrr6By2/7D7c8+m7c8WFefedLnpm/IiHOhR+vjRmqi0lkEuJwniX/8tv+k7AtLQ5NccGNL3D1HQui9VjmwxDWGJzq4cxjVrr8mReFr1+/fvTr1y/oMIQH0nb+y2geLmOY8Pm4bfdZvr4buDuTa7gVjvkL3H6fXXv2Jxai+2Dt1zuo6Nw2YXuqucCs7OdBsr8BO7HWSFnXW/zjw+/a7Z7ArN9e+HENCz+u4eBu7fhyXa23Ocgsb73UeG92PR/LV2/hpofe5p5fn8oBbcrSv47BqXeusYYr9tqPvfQpT837jDuuHkrPrrGzKYfDYcuon/GF8X7qG8KRBLIO26FbgI8+28idT0SS33OG9Ylu37l7f0JbWMNK6OFKMtt9rcdEuPHczj9L1Rt2MOCQjjHbYhMu+2PNerR/vvgp55+uMopPFIaxY8cGHYLwSNrOfwUz07z1huBU8/Lnf7zn+3UX6w1cess829nii0Kh6F/9qZj1L1/WbGf+InPur7idUnQQZToDeFHc8V+u877EjvVcZjJpNzXDM/NXsLV2L++7HDp0vJ5DgblTDZfZ87i1dm+0Zs4UpnE6iOhhlqFJ88nCKX9vXArI+sSi0zxUdolKsqG6ZD1c2Rb/s2CNvc7lFBtCCCEaFUzCFT/TvJ1VNemvEai+cVD066+3JK4hZxYVr6hOHEYJhdzXk5lJyX3PfOS4T6ozxd8kg2Sdv6rYMgQX76ADWgKNawt6ZSZc8Z+30+LV1u8SEi5LD5fd+ewmK927v/EcTk/x2b3/cNIhRfvz+MF6arvrxG+K+YPGoQvZesxT8zJb41EIIQpNASVcqXu4Nm3bw11PvJ9Wj4H1ScG7nnzf5rqRf1dUb7V9ze1NM1oXlOSR+pTJW4b5ltPZvTyEYO0FifZwNZhP/oWZ88YX7Ni1j1YtIqPae/bVR5NXL0KWIT8rx2khLN/GDzf+4Pp/R3upVq41YrLM82U3lYM1aXP6uOymS0hWe+j0ua/b5Pzgg1d3PfF+zNQnvbpFhllbtYj8/CebdiXKsnnGy/J0k4DJkyczefLkoMMQHkjb+S+jGq5c4uYpKoBX3lnNuO/0d31ea6+E3Q3TvCk6rc3o9pF+Mymps8Sebp6Taf+W0w2+IQzprkDT0BBm9uuf8/bSdZw4sBxoHHar3rCDac9+zFtLajisVwcgErs54asX0RquuF4kp7UUzfeaqlPwyVeXM+7Mxp+X+oZwwtQTEEkY488NkTqvlmXFPPDcEl5/P3Fh9XCSRMap+Z//36rkQSexbtNONm3bk/Cz9co7q6PLDQHc98zHQGQS4b3767n6jgWOcdqRubiEECJWwSRcmc7D9eDsJZx8ZAVtWpXy2ntr+NF3+hMKhWLOGz8ktGnbbtueLae4kjGHqayJQbpDTJmOQDneR8NhrOmc/nIz5Z3aJi1yr69vYPpzkSfcvnl0ZAEC80ZtDv8tX70FdXDjkG2yOalSMevX4m/09Q49XOZ3IUJp/bw0NITZvz9xegvrz4Y14R97w/OcfGRF7LQfceeLfu3wlGKyY5x8Xr2V+oYw/SxD4gDX3/tfNm7dzfUXHZfyHBDpqVyxZmvMAwKPvbjMdt/4ZZCqN9TSo0s7231F8zBp0qSgQxAeSdv5r2CGFK03q8+rtzFq4nOOvU7xduzax7MLPmfyA29x66Pv8vRrn0UXQbaeV3+5JeaR/p9OeZU3U8zB5PZmXmRT5/TXuNnp4890z9MfxtQOhTMsrk7Ww2X1q7ve4Lp73kh6rpstBeVmgXrjFA2N7/EDY91ISD5/03vL1vPTKa84zuVl9/nFXLMhzJd2c2aF0utJbGgIc2DbFgnbkxWVOyVbEPvzER+HU1xuhnivumMBE+98nYaGMM+9/nn0CdaNW3cnnDvZepGEEoe5d++1b4P4sDJ56EIIIQpNQfZwLfp0PQBvL12X8Hh7ZOfYb83k6sC2jZNabtm+l24d2yT0Jvzm7je577rTAJcTPLq8mZvxOz0B9sL/VibMPP/iwlWs37STP1x6EuA+uQuHwyz6NPGpwPg5rKz7x1uzfgcA7yxNveSRuVhztIDdOF0oBJ+tifQQ7t5bx9/+5fzAwH2zPmL95l1s2raHbh3bJLxullXFz3JufdLvCmPuqjlVo2PaJZ0atYZwmC4HtWbtxp2ccnR3XjcWza6vb6Dq8UXRJ0xdn89hWogHZy+JDsUmHuM+3v99vJYHnlvChi27GD96oKdzeJ05Pnce4WiklOoJPAp0AxqAaVrrO5VS/weMB8y/AH5rTHuDUup64BKgHrhSa50467AQQqRQQAlXYg+D22mw9u2P3FDKSoujCZe5nmH8zfirr3ekFZfbJGjJ55s4vHdHxyTuXodk5H1LD5HbNRAXflzDnx5pnA/q3qc/ZMgx3R1jTfYO0pnOIbGeqrGB/vWfxElCraJrJaZIFOrq7J9StPZMQmOvToj0erjqG8LRY81ky9yebrIF8fPHNX797ILPeWtJje0xTu30wsJV7N9fz1mnHBrdZhbzv/7+V8x+/YvGeGO7uByFgH//d2WSdwAr1mzl6r8soHvn2EQ4hx6ataoDJmqtFyul2gGLlFKvGK/dobW+3bqzUupwIqtoDAAqgFeVUv201ukvm9AMzZgxA5A5nfKRtJ3/CibhaohJuCI3dKeJRz9asdHxWPMGNfmBt7j7V99yrmtyyW3vycxXNLv31tkWd7udX8ttcvf11tjpLV5YuIoXFq5y3D/ToUqTmVCkW2M35obn2WnUENkd29DQ2GMXP+RoXjO+xso8jXXqjk4HtrSdwd/ufG63p5KshsvxWg6XuvfpDwFiEi7zZ8dcmsgU+1Svc3xbavfy3rL1zjsAC96PJJpffR379OR/FlVTeVhXykqL7Q4LhNa6Bqgxvq5VSi0Duic5ZDQwU2u9F1iplFoBHA+4WwKhmVu+XKYHyVfSdv4rmBouu+J2p4Tr9sdiF6I2bz7hcOx8Sp+s2uyYMMX3mDjFlE5usXrd9oReqnTri/zcL7q/JQinST3Tua5Z7+S2B2SnpWC7wajFss7b9fz/VkaHheOHZM23an3L+yxt/Pv7F0Y/426dEocq7d6DXZt4WbYoEpfzU4pOP7/pDIE6fcax82plllA7vfeFH9cwy2Z5o1yhlOoFHAO8bWy6Qin1kVLqIaWU+bRBd8A6q3E1yRM0YTFmzBjGjBkTdBjCA2k7/xVOD5e1aNko8i12eUf/2NLjZZ0NPYTzzWimi3mGDjqgRVq9OaFQKKGAORz3hKCdhobIUW5vxOneYNdu3Mk7S9cxdrjightfSOvYmOv6MHN6OAxX3P4f2rRqXKJp3abGRcKdiuat2/fur49JmswapWQTx5qvOLWn0/qCqcQkXHGxO/Vsxn9+e/bW0bKF/X/KTudwM2+dW8nmjrNbXSAXKKXaAv8CrtJab1dK/Q24icgA601AFfAT7P/jy803lYOUkiWe8pW0nf8KqIfLpobLZRHX40by1BAO28wsbn+Mm3P36NI2vbm0Qol1WBNumZfysKv/soCzfz3b9bXSvcFOfuAtZrysYxIbL+KX2fFS4mOew9rrZc0p4qeFMPe3JgV799XHJLbmHFNu2rTWadkej4mFtSkSergcwonf7/8eeMvx/P94wX4Kh/2WeDPt4Ur28MjMVzRVjy9yfD0ISqlSIsnWY1rrWQBa6/Va63qtdQMwnciwIUR6tHpaDu8BuFsNXgghLAoo4Wr8OtWQYjLWZCTZ0jz2i03HnashzeEfIP6P5w2bUyc55rQLXTu2dnedNKuZ9+yNTCkQXwe0btNOPv8qcUkjJ9HFmDPs4Ypn7Zmy3vxf+N/KaMzWJGWvzTxa8edJuK7x773/+iihBhBAr96SLGxHyRaFduzhivsQkk1/4vTzY9Z7gbeVBKxSPcXo5WGCbFFKhYAHgWVa66mW7dZHQs8BlhhfzwbGKKVaKKV6A32BdxCuLFq0iEWLcivhFu5I2/mvcIYUw4k31HTXFgyH45OBkGNyYLeeXuL57Ot9nMRPtJquzu1budov3Rus3bAcwPg/vurpPNHEwsNjbHbtYT2NNUanJzv37qu3T9w8JOimx1781NNxyVZIcBwO9HkSd6/1Z6ZkQ4o56GRgHPCxUsqc6O63wFil1NFEcutVwKUAWuulSqkngU+IPOF4uTyh6N7cuXMBqKysDDgSkS5pO/8VTMIVM9N83IzmbVuVxsyUneQsMUOIoSSTYhYXpe7hChN7Q929ty66dmCyY7xye+NM9wbbWHieeXH1S2+t4t1Pkj/1lozdkjHWRKm+Psya9Zs63DwAACAASURBVLVJp+/w0sOVrRkOnnil8Umg+EW0rWs2Wp9WzbRHKt6zCz7P6PhME7ampLV+E/vmfD7JMVOAKVkLqoANGjQo6BCER9J2/iuYhMsuGTBvxG5vUIk9XM5JhtseLuvEoE/NW86PvnN46iBszuOG2/uwqwlb7a6f4X119framCkGvCQxe/cnX2+xrr6Bn//5tRTnqLf9TJ1y6HWbdrKlNvVTqZnavTf2vVkTwI8+28hR/ToDcXNo5QBZN1E4GTVqVNAhCI+k7fxXQDVcNjdQc7LMdKZWsJynKMMarkefX8bsNxonm7Sb3d0qFLKP1W2C5LYH6qlX05tfJVrs7vL8Km79PpOberRU9uxL7J1yquFyMmnaQj5ZuTnxPA5DiukOnXq1a09sL2zI8iO2a2/ja/FDeEFPMBqfKAohhEhUQAlX4rbMe7ica7i8+CJFgXkoMu95wvb4oSYncyzJXTLpvqXEGeJTcJkAuBvmjWX3WVhrndZsSG8lAKt0a/78VvX44pjvre/L+iMcn+AEPaG72zVLRfNTW1tLba2sqZmPpO38V9hDiiHzNbdncT8thJPundsmrR/63m/mOL7mVDNm16vjRtcOraMTgvrBqQg9XjYTALvPwponZdKLlknRfDZYw/nLzMV8+uUWFn+6PmFy1/gfmSWfJz5FKUQQpk6NPAg6adKkgCMR6ZK281/BJFy79yQOa5QYw35ue6nie7iSTQvhdGvu3L5V0oRrX7JJIhvCtgnFnn3uh2xKikPRG7LflT4b45YEcpLutBPpsKvh8ut6mSRcPbu2Y836zP4a7N65TczyONYet91763nGYdb2+B/R6+/9b0ZxJFNU5G+vryhsbdu2DToE4ZG0nf8KJuF6YPaShG13PfkBpw8+2P2QIrh+SjHTJ/bsOK1Zt22H++V0SoqLqKs3kra4GAvhZrlnr10Nlz/ntiY455/eL+YJwlRalGY+Oh+/FmGu9bgBtCgtlpot4drEiRODDkF4JG3nv4Kp4SorcX4rrh+jt9nN6ZF3x0fhs3CPvO6eN13vW2r5HOIjLE3yGfnJ72kLrGxnTvfpM7dOw+B2WShTNp7US6fnbnGKBzL80rIsdxaiFkKIfFIwCVf7di3p0+PAhO3hcNh1HVb8OoYQckwenJZyCbpPosTy9GR86K6L3jPU1H1ofhW7W3uUSkvTSyx2xfX6FPvQO5VOMfqk6Qszvp4bTTE9hhBCFKKCSbgaGsK2QzCZDKFt3r4n7cWKs1m/5Eay6Sqaajgx0x6u877dz6dI0mNtuxZpJlxfb4mtbzuwbQtm336WL3EJka+mTZvGtGnTgg5DeCBt57+CqeFqaAjb9nSkWufNKj5PeOTfnzjsF+bDz75OK76mUmpNuDJ84tKrTK9z4YjDaNe6lAdnL3W1v289XCH7rz2dqygUePItRNBqamqCDiEQW7bv4aADWmbl3Pv211OW5h+EXjTXtsumwkm4wvY9XOms8+Y2T/jfRzUsX73V/sWA77HFlhnwA6uPd9nDdeqxPXntvTW0blnCLstTpqFQesnKww6Jsald61Jqd6We88s63UIojYzL7mGEHKx3z4jZVkKkY/z48UGHEIgfTX4pa+eeUzU6a+e2aq5tl00FMaQYDof5aMVG2yfY0ipmdpkorNng/Ph/0PdZM+k8pPuBZFJN1b9XB8/HDuzT2dV+36rsAdivS+mlc6hPz/YJ247u25kJZw90dbx1lvpQKMQRh3Z0dZzdUG0uPmGYSrLP/KwhhzCnajR/uXpo0wUk8l5FRQUVFRVBhyE8kLbzX0EkXObknl+sTZzJPZ2Ey+2++xwWP4amLxiPVxQKccfVQ5nys5Ndr61op8TF0kXxWpQVM6dqNAd3a+dq/9KSYuNa/iQnB7QuS9h29Q8HuS6Aj1/W6bj+3TzHko/DifExDx7Q+P5bGE8nHtqjPWNOV00alxBCFIKCSLiS3dzSqeHamySRitkv2czvAWdc4XCYPj3a07ZVaUIC2a51qevzZJIvuC2aNxMtu0J/L8miXa9ShwNaun9i0HLNolDI02dw5km9osdbqYPt15eM1+8bib10fknV/vEf04RzGnsGW5Q2Vh+MHa74/ql9fY1NFKb58+czf/78oMMQHkjb+a8gEq5kN9R0erjcrlmYLDHL5hxUbny5rnG4M36twnOG9WHkN3u7Oo+XZCPdt56sh8vLx+j0c+A24bIODRYXF3nqpTIfWogfJXU7B1rXDm3SvqZ7yd9P/PttWdaYZLWwzL9VVBSidcuCKf8UWbRgwQIWLFgQdBjCA2k7/xVGwmW5YfeNq+PZX+d+HULHyUzjZGOSy6ZQXFTEtyp7uto3k7zRPNZuXrSYeMweLtuEKP0Aih2GJpNNlWFlHVJs07LEWy+fcUx8D1fIZXVfcVGImy89ycOFU7MplYsRH6E1EW4hE54KD4YOHcrQoVL3l4+k7fxXGAmX5U5y82WxN6uvNjiva5gNQddwJVNU5M+EnI7CsWs49q5InnCZSYldcucl4SuJyyiO7d8VcP+erdds06rUU8JlJlbxvUVuz1VUFOKofu4eOkhXqukzjuwbe12zBxKSr+QghJNhw4YxbNiwoMMQHkjb+a8gfotab6itW8bWqVQ9vhiAH488POG4U49119sT78t1251fzOWMi5Dr3h5PQ4rmvy6zJbPdkiVc7VqXcdOlJ7o7X1wP17GHdYm5TioxPVytSl33SlmZn5vXpxSzmRCnatPRpxwS8721hysfHwIQQohcklHCpZQaoZTSSqkVSqnrkux3nFKqXin1/Uyu58TNTaqic+LK561beKtD+bw68WlIU+LyQLkjFHI/P5SXHqb4Y1LepI2X7T+zyLbTjutJz67unnpsGdee5lxabp+4tNZwtSzzOKRoiP+c3U7O6jYhTtcfJpyYsj2sSeIpR3dPun/ApYoiT6xdu5a1a9cGHYbwQNrOf55/uyulioF7gDOBw4GxSqmEbiRjv1uBrM0E51S7Y2V3042/QTu599pT047pyvOOTvuYppCtG3qEMaTo8mZs9iAl6+EKhUKuk5X4BNo8ym1vkzWOslL3RfNnDWnsGTKPSbimy+QtWz1c5Z3apE64stSL5bUnWeS/6dOnM3369KDDEB5I2/kvk7vv8cAKrfUXWut9wEzAbgrcXwD/AjZkcK2k4ifOfPjG4Qn72N1029nM22SnZ9d20Uk6UzFv2tYnvIJmJgThcJZruAxm3pKygyvaw5XkHLhPmNq0ih1ONhMML0OKpSXFrnq4Kg/rEtMDF03y4g5OlsxYl+lwG+s5w/rYbnf6OQ25mOYiJiFLsW86PbkHtHH335koPOXl5ZSXl3s+Ptmch5nasn1P1s5dCDJtO5Eok6ygO2Bd66MaGGzdQSnVHTgHOBU4LoNrJRV/Q25rk0jZTT1QWlLEsEE9mL+4OuU10q1hKSvNofK4aOj2yx/ZcVq0OxnzELOGK9VnFooWzSdeyzyy3mFRcjsD+3SyPb+1d/Oyc4/kvlkf2R5vXbC6tCS2h+uPPz+Z3977X8f30Ph98u12zhl2KE+8shxwn1z+ZNQAnpm/ImF7lw6tbfcPhVK3hzXZ81K/JkS8CRMmZHR8WWkxoyY+51M0Ih2Ztp1IlElW4OZZ/r8Av9FaZ+/PFBt29yyn5WN+9r0jE7Z3PND7oqNm7uDH0i7P/nlU0tfvcLnMitm70tDgbU4qr0LA738yOGH7XROH8asLKqM/QHa5XY8ukZq7ffvrXQ919et5EEOPaezhsStg79k1sZbPZJ3Ms6ykKOYHvIPDQrROocW3f9Jkx/L+05nh/9FJZyRss/s5h8jPQKqmj+ngMr4+um9nY5moOGn8eGRrqFIIIfJJJglXNWAtzugBxFfYHQvMVEqtAr4P3KuUOjuDa7qU+Avers7L6TYw6acnJO6b5J7RqkVjz4g51OLHU12p6q3Sv5GFY27I11/k3Onodk6y2LMb/1oO/YbNMj+9Kw5k6KAeJMu4jju8G5ecNYALRhzmOnkNheIm6DQOsyYxdu3S0jjGemxk4lPLuR2umTDfllnD5XFaiHSGfA+ySwIdeiZDodS9VnZDijdddhJ3XjPMVTxO71HyLSGEyCzhehfoq5TqrZQqA8YAs607aK17a617aa17AU8DP9daP5vBNV2x7+Gy2RgK2d6AW9isved0sxp4aCduv/IUu1NnnftEpLE43XrMSUc6L0zqqZfPuNmbayn2790haeIZnYfL7rWiEGcP7cOBbVvYvs+fjj4iYVsoFIoZnrSt4bK5mHONUerPN/7tNfaqxW5PlhxbQ8q0xs7uaVxTqp/JopghxeSsMf945AAgdt6u2OtKxtVcVVVVUVVVFXQYwgNpO/95Tri01nXAFUSePlwGPKm1XqqUukwpdZlfAXph9wvebqgm8ld/o6f++F1m3TrS9nine8Yff34y3+h2QPR769N1dqZelZiceZXqPvbHn5/MdT86rrEjicQb+sk2SdeUn53E8YcnX7i5vGNkCRrrbPLmTfiIQzvx4A2n863KnkljPKhdC/r0OJArzzsm6bXskpX4JMmsXbJ28EQTLkvv5q49scsdAUz52cmceWIvyjvFL6uTupcvsVYrZLvdrTVpTtRr1ibOuPk7zLp1pOOQZH1D2DGmTkZybf2c04nfrFcceXJv29eXrdrs+lyisOzYsYMdO5p28mnhD2k7/2X0KJ3W+nng+bht9znse3Em10rljBMO5ihjpmy7e4VdL0ko+n8RxcVFRrG09zjMm5bTKQ62JGdzqkZnVBBql4icOLDxqZKBh0aKyD9bswWIFKfHJ1zXXXRcQgxH9unMq++sTnpt8ybbt+dBXHn+MVxZNT/m9S4HRRKgZMNYxcVF3HH1MCBS13Vl1XwOtVkOyLHtLM41ntpriOnhMq5j6W6q3ZWYcHXr2Iaff/8oIFIXt2rtduNcjfuUlhRzQJsytu/c5/h+rHHFx5ysN9LaK7d+807H/Uae3Ju5/10Zs+32K0/hrSXraGs8oem0fE9dXQMXnnkYtz76HoMHdOPtpeuAyH83i/WGlDEmU1pSxKxbR1FSHGKWTSH/+s27PJ1X5L9rrrkm6BCER9J2/suhR+kyc8UPjmbI0d0B9z1c8UOKye43bv/iN3tTnPpGMi2mv2DEYUY89uf67cXHJ2yzxu5mzrII596dy84ZyPATDo7GYR2yjJdq/T5T74oDuXbcsUwenzirfPzbPPXYnglZtfltg92QouU9d+nQKmkcfXq059vHfyNyLiPj6nhgSzof1Io//2KITWxx0y2EGrfbxZdKstI5u5qtQ3u0j/5M2F3Xeuw3j+rOnKrRHNarQ3T7BWccZumVa9z/hCOS93DG9yTGP9VpdYDL6VdE4WnXrh3t2rmbuFjkFmk7/xVMwpWKXW1MiNiekmTDQekWPdfXN3Cbww06lT/9/ORoQfuZJ/Ximh8Oir52xuCDo9dxWzRvTUaKbDKg268cwvnf7hezLdmsEMcPKI/2XCUkHAnXdp9gDjm6Owe2bZHyHFePHcTuvXX2+1jiNpvc2vZH9nG/TqHZ8zR4QCT56G5XH2UJbfjgg2M+l9jdXH4OST54N0sm2SXhB7VrQSvLpLB79lk+O8uwujXmEwc61/clXDPFW6tryM/F3oUQwk/NKOEqYtatI2O2RXpnYr+3/puOO64aysQfDor2pNXVh2N6EkxueriOOLRTtKD95987im9VWh4GjRZlFyWca/QphyY/cdj+5qgO7sCFZ/aP39VRSXEo2pMUSpH4ZatcevvOvTHfF1mSysZrmz1c9j/mP//ekdxy+Tcdr2GeK9n7KwqF4np7Yv+1BONKsh6uunpvCVf8+9+2w35YNK3e15jkz/44s76srk4SruZqzpw5zJkzJ+gwhAfSdv5rPglXccjmKSqHgmebG0iqnpo+PdszrLJn9KaVah6rs4dGkqNrxx0bs/3ofsl7YMybf3FRYs+S3ZN71mMawu57nJJ1phQXF1FS1HgzddvDdclZR3DzZSe5un4qPbvYd3XHJD/GT7fTk39nntSbAYd0THktu4/iPKNHMOEpRePfhBout1l8ks+93kVPkd17jd825nRLb2Y4sz80IHbYuCxm8tjI13X1knA1V4sXL2bx4sVBhyE8kLbzX+6sP5NltkOK8VXzMdvjtrm8jtnDlezmOKeqcQWkIUd358//eC/6/cQfVrq6TnFRyP0UAuZuacweHz98ddk5A7nvmY+BSM9Fa6NIe+ee/Ul7RqyfpZlkZuLw3pFew/gpLRqfDrVeuzE59cJc+ql2V2KPkDkxa4j4Gq7Ya53/7X488erypNexftSd2jvXmHnt4YpfZaHjga3o1L4VG7fuZl9dQ+MfGGlMvRbTv2V9zzHLIxWxe68kXM3ZyJEjU+8kcpK0nf+aTQ+X05NutsmV3f3Z2Ga9ebWyWfzavLm7uTnaSdXLYN7Piovt5xCzPae5SHQacVj3PXFgOd/9ZuMCzcXFRdGFonftqUs+pOjzHEy3XpFYFweNCaK1nc0vvcZg1pPttxkSiyZ4cf8FRefhMr648Mz+zKka7ar+CuCK846K+d5M7CBSF5iKed3De3eg6peRKUjs6vbMxLWkOOT+rwkH1s/X+i5LSyLXtfv8RPNQWVlJZaW7PyJFbpG281+zSbhsJzN1uNdYbyDmTcvc8k3jSUiwX1qnyFI0n65rLzzWtmjcyuw5iy+an3D2QMdjzMQjrdnjjQThjBMOTnjysaQoxMA+nejTsz3jzuyfdDmaJlgrG2isfYpN/mIvbi7i7dZRfTvzvW/14dJzEj/bsEN9Vyjhiwi3n338gurWJyP3u/iZMnvIBhzSMdouduuIXnn+Mdx25RA6HtgqZp42t2KGbmO2N77QrnUZR/ftzK8vjB02F0KI5qhZDCk++LvTY2pLGoVsu5SsW/p946DINpv5teyeWmscUky/h2vIMd1T7mPe4Du1bxUT+qgkyUTXjpE5seyGxqxuvGQwm7dHitHN6O16h4qKQrRuWcodV0USzm079ibsE9VEs4ybNXMxCzBbLm0dxnWruCjExcYs6vEce6zierhMyXp5rOeKP86agNW76DXt1rEN919/Gl07tOHz6q2A/ZBqi9JiDjs40ssVfYrV4/qZ1pitH0tRKMRNPtXsifyktQZAKRVwJCJd0nb+K9iEKxRq/OVfWmrfA+Omhyu6LclrVqcd15OX3/6SI/t2SiNa9w46oCVXjTmGQYd1iW5r3y55r9iQo3uwcu32lDVUx1lml7eriTLFfwb2yWxENnu4zjjhYF5660ugsecvdkjR/4v//pLB1NeH2WEkr0WhUDQpP/7wrny5rhZI/LnaX9e4frvdBKbRmJN8YKOGHMLY4YqduxMnb7Wq6BT5Q6AxCU3VkZ3+5xS29IdZh1VjUrYm6t0UuWvmzJkATJo0KeBIRLqk7fxXsAnXU38ayfevmws4r/EWDjslFDY7u3yS6/DeHR17U84xZkPP1GnHRSbmDIfDfO9bfaLfOyktKeKSsxqfYLzp0hOjS+E46VUemRH/cJupLeKVlQQzMv3t478RTbjMmjmnHi6/mEsemdcNhSKTtj7z51GUFBexen2tcfHY46w9XId0T5xN3w2zTZIV1lsd2uNAju3flR99p3/S/Xp2bctXX++IWbw7HdYEfNyZ/Xnk358ATTecLHJXv379Uu8kcpK0nf8KNuFqEfN4un1CEA47ry8Xz9wv3V6TDge04NAe7bnxkhPSOi7efdedxvpNsUukhELOQ17JHN2vS8p9BhzSkYduGE7ng1Lf3J3muYLs9DJFWbpTzJo5ayypkspMVBo9jN85qTfQOJTsNLWImXANH3yw60SwwwHJey5TKS0pZtJPU//cXT12EJ+u2hJdjskVy2ffoV3jDPjfP7UvR/bpxMQ7X2+y4WSRu8aOHRt0CMIjaTv/FWzCZeVU1O1UspJs7b50n7h6ZNKItPZ30r1zW/uZzrPITbKVSiiL3RzWeiGzoPycoX1YsWYr488+ImbdSr91at/KticzOvQc9yNn/tx89+Te0dqqZB684XRat4xMvXHmib3o27N9RvEm07placwQdTqO7NOJ/r1T94IKIURz1ywSLqd5mJwKn1uW2Xwsxin27o/U4vR3MdRWaG669ERWVG9L65hs9nFY64jMgnKnNQ+bSnQSUYfX7Xpb7X4Mrb1N5sLaucQMeWCfxFrFtq0jiWKfHtlLEoUQIt80i4TLidMzWXY3RXOIyEy4yhwK8QvZ0f26OA5H/nT0ERxSkVib5Nc8XH++Yghb456GtCYquTO5pjGkGPe+r7/oOF5660t6dGnLp6s22x454sRe2Q4updt+MYRN2/a43t+udSs6teX2K4d4rlUThWPy5MmAFF7nI2k7/zXrhKs0Se3RWUMO4RuWISnz/nnEIR3peEBLxgyXR2WtnNZx9GtEMdWwVa6s1xc/8amponNbfjwqUm8Xn+ib33fLYs2ZW3brf9pJNZGrOrj59QALIUQyzTrhKknydN14h4lEi4uKuPL8Y3yNo2uH1qzfvCv1jvkoi4XT1oWq/ag380Mm6xLmZY15PsYsmoz0juQvaTv/FXTC9cDvTmfj1t2Orzs9vWjHHCJyu0RLOu6aOIydu+t8P28uyOrUAJamGD3Unyk3MmUWx2/fmXySWauzhx7K59Vb+fbxB2crLGFQSvUEHgW6AQ3ANK31nUqpDsATQC9gFXCe1nqLccz1wCVAPXCl1vqlAEIXQuS5gi5E6tqhNQMO6ej4erIhxXhelj9xq3XL0pzpofGb32sp2hlwSEfPC1T7bcnnmwB4e+k6x326xk2/0OGAlkz52ckc0KbM4YjcFT/9RR6oAyZqrfsDJwCXK6UOB64D5mmt+wLzjO8xXhsDDABGAPcqpbxNWCaEaNYKOuFKJdmQYry8HO4pcN06tgHgm0dVBBxJo+3GDPTXjnNeP/Cofp25/crgnqT0g7nmZ74liVrrGq31YuPrWmAZ0B0YDTxi7PYIcLbx9WhgptZ6r9Z6JbACiF1cVDiaMWMGM2bMCDoM4YG0nf8Kekgxlc4uZ+y2ysaQovCm80GteOpP37VdmDyZmy49kYYs1diPO7M/dz/1Accd3jXpfvleVD7ym4fQrnUpwwb1DDoUz5RSvYBjgLeBrlrrGogkZUop83Hc7sBblsOqjW3CheXLlwcdgvBI2s5/zTbhuvfaU+nZtZ3r/RtruLIVkfDCds60FNzMtO/VINWFh24YnrXz54riohCnHpt8SalcppRqC/wLuEprvT3JAr12fdvyW8ClMWPGBB2C8Ejazn/NNuFKJ9mCxiHFsPyuTdvB3doxakjyhbOFaCpKqVIiydZjWutZxub1Sqlyo3erHNhgbK8GrN14PYC1TRdtfkuSyIocJ23nv2abcImmc/evTw06BCEAUEqFgAeBZVrrqZaXZgMXAbcY/z5n2f64UmoqUAH0Bd5puoiFEIWi2SVct185hMWfbki9Y5xQYxeXECJ/nQyMAz5WSn1gbPstkUTrSaXUJcBq4AcAWuulSqkngU+IPOF4uda6vunDzk+LFi0CoLKyMuBIRLqk7fzX7BIudXAHTwXLZiGH04LXQojcp7V+E+fpWk9zOGYKMCVrQRWwuXPnAnLTzkfSdv5rdgmXZ9GJuCTjEv4YPKCbpydlhcgXgwYNCjoE4ZG0nf8k4XIpDyd4FDnuhp8MDjoEIbJq1KhRQYcgPJK281+znvg0HVLCJYQQQgivJOFyqbGGS1IuIYRwo7a2ltra2qDDEB5I2/lPEi63ZG0fIYRIy9SpU5k6dWrqHUXOkbbzn9RwpUs6uIQQwpW2bdsGHYLwSNrOf5JwuVQkNVxCCJGWiRMnBh2C8Ejazn8ypOiWmXBJDZcQQggh0pRRD5dSagRwJ1AMPKC1viXu9dHATUADkVmarzImHsw5Jx9VkfR1mRZCCCGEEF55TriUUsXAPcDpRBZ4fVcpNVtr/Yllt3nAbK11WCl1JPAkcFgmAWfDnKrRrnuupINLCCHcmTZtGgATJkwIOBKRLmk7/2XSw3U8sEJr/QWAUmomMJrImmMAaK13WPZvQw6XQIVSPIUYkiFFIYRIS01NTdAhCI+k7fyXScLVHVhj+b4aSJg6Wyl1DvAnoAvw3QyuFygzIZN8Swgh3Bk/fnzQIQiPpO38l0nCZdcllJCOaK2fAZ5RSp1CpJ7r2xlcMzBSwSWEEOmpqEheGytyl7Sd/zJ5SrEa6Gn5vgew1mlnrfXrwKFKqU4ZXDNw0sElhBBCiHRl0sP1LtBXKdUb+AoYA/zQuoNSqg/wuVE0PwgoAzZlcM3ASA2XEEKkZ/78+QAMGzYs0DhE+qTt/Oe5h0trXQdcAbwELAOe1FovVUpdppS6zNjte8ASpdQHRJ5oPF9rnZcZi9RwCSFEehYsWMCCBQuCDkN4IG3nv4zm4dJaPw88H7ftPsvXtwK3ZnKNXCFLKQohRHqGDh0adAjCI2k7/8nSPi51at8KgC4dWgUciRBC5AcZjspf0nb+k4TLpZOPrGDy+BM5ul/noEMRQgghRJ6RhMulUCjEoMO6BB2GEELkjbVrIw+uyxQD+Ufazn+yeLUQQoismD59OtOnTw86DOGBtJ3/pIdLCCFEVpSXlwcdgvBI2s5/knAJIYTICln4OH9J2/kvlxOuYoB169YFHYcQoolY/nsvdnuMUuoJ4GXgMa31nmzEJYQQmcrlhKsc4IILLgg6DiFE0ysHPnezo9b6fKXUcOBxpdQK4E6t9VdZjU4IIdKUywnXu8AQoAaoDzgWIUTTKCaSbL3r9gClVCUwCvgCmAtUEVlqTASsqqoKgIkTJwYciUiXtJ3/cjbh0lrvBd4MOg4hRJNz1bNlMQy4QWu9DUAplZfrtRaiHTt2BB2C8Ejazn85m3AJIYRLeyzJ1mXW5cVEsK655pqgAOQJWwAAIABJREFUQxAeSdv5TxIuIUS+O8Tyde/AohAJ2rVrF3QIwiNpO/9JwiWEyHdhpdQkIAzIMvNCiJwkCZcQIq9prX+llDocCGutlwUdj2g0Z84cAEaNGhVwJCJd0nb+k6V9hBB5TSn1M+B3wI1KqceDjkc0Wrx4MYsXLw46DOGBtJ3/pIdLCJHvummtZcK+HDRy5MigQxAeSdv5TxIuIUS+q1BKnQPsBNBavxxwPMJQWVkZdAjCI2k7/xVEwqWUGgHcSWTSxAe01rcEHFKUUqon8CjQDWgApmmt71RKdQCeAHoBq4DztNZbjGOuBy4hMuHrlVrrlwIIHSOWYuA94Cut9ch8iFsp1R54ADiCSCH1TwBN7sd9NfBTIjF/DPwYaE0Oxq2UeggYCWzQWh9hbEv7Z8OYtPRhoBXwPPBLrXU4zXDeBNoBBxD57IQQIufkfQ2XkRDcA5wJHA6MNQpoc0UdMFFr3R84AbjciO86YJ7Wui8wz/ge47UxwABgBHCv8R6D8kvAWoicD3HfCbyotT4MOIpI/Dkdt1KqO3AlcKyRwBQbceVq3A8b17XyEuvfgAlAX+N/8ed04zOgg9b6EWCrh+NFlmit0VoHHYbwQNrOf3mfcAHHAyu01l9orfcBM4HRAccUpbWu0VovNr6uJXLz704kxkeM3R4Bzja+Hg3M1Frv1VqvBFYQeY9NTinVA/gukd4iU07HrZQ6ADgFeBBAa71Pa72VHI/bUAK0UkqVEOnZWkuOxq21fh3YHLc5rViVUuXAAVrrhUav1qOWY9JxMdDe+PoUD8eLLJk5cyYzZ84MOgzhgbSd/woh4eoOrLF8X21syzlKqV7AMcDbQFetdQ1EkjKgi7FbLr2fvwDXEhkKNeV63IcAXwN/V0q9r5R6QCnVhhyP21hs+XZgNZH1Q7cZtUg5HXecdGPtbnwdvz1dO4EGpVQI6ODheJEl/fr1o1+/fkGHITyQtvNfISRcdhMd5lwdh1KqLfAv4Cqt9fYku+bE+1FKmfU5i1wekhNxE+klGgT8TWt9DJGb8XVJ9s+JuJVSBxHpCeoNVABtlFIXJjkkJ+J2ySlWv97DXCJt/gwww8PxIkvGjh3L2LFjgw5DeCBt579CSLiqgZ6W73sQGYrJGUqpUiLJ1mNa61nG5vXGkArGvxuM7bnyfk4GzlJKrSIyTHuqUuqf5H7c1UC11vpt4/unidyMcz3ubwMrtdZfa633A7OAk8j9uK3SjbXa+Dp+e7rWAL8x/velh+OFECLrCiHhehfoq5TqrZQqI1KcOzvgmKKMYY4HgWVa66mWl2YDFxlfXwQ8Z9k+RinVQinVm0gh8TtNFa9Ja3291rqH1roXkc/0Na31heR+3OuANUopZWw6DfiEHI+byFDiCUqp1sbPzGlE6v1yPW6rtGI1hh1rlVInGO/5R5Zj0jEWOJ/I06j3ZPIGhBAiW/J+WgitdZ1S6grgJSJPdj2ktV4acFhWJwPjgI+VUh8Y234L3AI8qZS6hMjN9gcAWuulSqkniSQJdcDlWuv6pg/bUT7E/QvgMSMB/4LI9ApF5HDcWuu3lVJPA4uNON4HpgFtczFupdQMYBjQSSlVDUzC28/Gz2icFuIF439p0VpPtsR1jce3JLJg8uRI00yaNCngSES6pO38FwqHc7XsQwghUjOSP/MX2Xqt9dVBxgPRB2RWzps3jx49eqTavWD5cdMeNdFLp6f/5k6NPEA78ppnA41jTlXTPIQvCVf6qqurOe200wB6a61Xxb+e9z1cQojmTWvturLXYcLW/wPGE3m6FeC3WuvnjddyYlLcfCU36/wlbec/SbiEEHlNKTUbKDW+3Q/s0Fr/0GH3h4G7icz5ZXWH1vr2uPNaJ2ytAF5VSvXLsSF+IUSeKISieSFE8/YRkV6rkcCHSZItpwlbneTSpLhCiDwnCZcQIt/1Bboa/zvU4zmuUEp9pJR6yJgTDXJzctm8MmPGDGbMkKnR8pG0nf9kSFEIke9+D5hPJ05OtqODvwE3ESm8vwmoIjLFRD5NLpuTli9fHnQIwiNpO/9JwiWEyHcdiUx2+xel1FlAWivuaq3Xm18rpaYTmbkecnNy2bwyZsyYoEMQHknb+U8SLiFEvvsxjYnQKaQ58bFSqtxcAxI4B1hifD0beFwpNZVI0XwuTC6bVxrnHxb5RtrOf5JwCSHynevFqx0mbB2mlDqayHDhKuBSCH5yWSFEYZGESwiR7+YSWV3gGVIs7eMwZ9eDSfafAkzJKLpmbNGiRQBUVlYGHIlIl7Sd/yThEkLkuwqt9dlBByESzZ0bKYeTm3b+kbbzX84mXEqpFsBxQA2RWZ6FEIWvGCgH3tVa73V5zDFKqU+BbQBaa3m8KkcMGjQo6BCER9J2/svZhItIsvVG0EEIIQIxBHgz1U5Kqe8QSbQmAsuI1GH9IbuhCbdGjRoVdAjCI2k7/+VywlUD8Nhjj9GtW7egYxFCNIF169ZxwQUXgPHfvwsjtdY/V0r9TWvtZQ4uIYRoErmccNUDdOvWjR49egQdixCiabktI/iGUmo40NP4F631y9kLS6SjtrYWgHbt2gUciUiXtJ3/ZGkfIUQ+e4pIzZf5r3SH55CpU6cyderUoMMQHkjb+S+Xe7iEECIprfUjQccgnLVt2zboEIRH0nb+k4RLCCFEVkycODHoEIRH0nb+kyFFIYQQQogsk4RLCCGEECLLUg4pKqUeAkYCG7TWR9i8HgLuBL4D7AIu1lovNl4bYbxWDDygtb7Fx9iFEELksGnTpgEwYcKEgCMR6ZK285+bHq6HgRFJXj8T6Gv8bwLwNwClVDGRdc3OBA4HxiqlDs8kWCGEEPmjpqaGmhq3U6qJXCJt57+UPVxa69eVUr2S7DIaeFRrHQbeUkq1V0qVA72AFVrrLwCUUjONfT/JOGohhBA5b/z48UGHIDyStvOfH08pdgfWWL6vNrbZbR/sw/WEEELkgYqKiqBDEB5J2/nPj6L5kM22cJLtQgghhBDNih89XNVAT8v3PYC1QJnDdiGEEM3A/PnzARg2bFigcYj0Sdv5z4+EazZwhVGjNRjYprWuUUp9DfRVSvUGvgLGAD/04XpCCCHywIIFCwC5aecjaTv/uZkWYgYwDOiklKoGJgGlAFrr+4DniUwJsYLItBA/Nl6rU0pdAbxEZFqIh7TWS/0MftmyZezfv58jjzzSz9MKIYTwwdChQ4MOQXgkbec/N08pjk3xehi43OG154kkZFmxbNkydu3alZWEKxyOlJuFQpFStIaGBoqKZJ5YIYRwS3pH8pe0nf/yei3FGTNmsG3bNubPn88DDzwQ3f7f//6X+++/n127djF8+HAmTJjAnj17+O1vf8uGDRsoKSnh4YcfZuHChdxxxx2Ew2HGjh3Lueeey3XXXUeLFi1YtWoVv/nNb7j55pvp1KkTAwcOlMdkhRBCCOFJXidcY8eOZdeuXVx44YUx2wcNGsSjjz5KOBzm/PPP50c/+hFPPPEERx55JBdffDENDQ0ATJ06lfvvv5+2bdty/vnn853vfAeAI444gsmTJ1NdXc2GDRt4+OGHKSsra/L3J4QQ+Wzt2shzUjLFQP6RtvNfQY6RffLJJ1x88cWMGzeONWvWsGnTJr744gsqKysBokODDQ0NdOjQgbKyMnr16sWGDRsAGDhwYPRchx12mCRbQgjhwfTp05k+fXrQYQgPpO38l9cJV0lJCfX19Qnbp0+fzg033MCjjz5KeXk54XCYQw89lMWLFwNEe7iKiorYvHkz+/btY9WqVXTp0iW63SR1W0II4U15eTnl5eVBhyE8kLbzX14PKR5zzDFce+21LFmyhNtuuy26ffjw4Vx99dX069eP1q1bA3Deeedx3XXX8fLLL9OyZUsefPBBrrnmGi699FIAxo0bR8uWLQN5H0IIUYhk4eP8JW3nv7xOuHr27MmMGTMStp977rmce+65Cdv/8pe/xHx/4okn8tRTT8Vsu+WWW6Jf9+jRg7vuusunaIUQQgjRXMl4mRBCCCFElknCJYQQIiuqqqqoqqoKOgzhgbSd//J6SFEIIUTu2rFjR9AhCI+k7fwnCZcQQoisuOaaa4IOQXgkbec/SbiEEEJkRbt27YIOQXgkbec/qeESQgghhMgySbiEEEJkxZw5c5gzZ07QYQgPpO38JwmXEEKIrFi8eHF0hQ+RX6Tt/Cc1XEIIIbJi5MiRQYcgPJK2858kXEIIIbKisrIy6BCER9J2/pMhRSGEEEKILJOESwghRFZordFaBx2G8EDazn+ScAkhhMiKmTNnMnPmzKDDEB5I2/lPariEEEJkRb9+/YIOQXgkbec/SbiEEEJkxdixY4MOQXgkbec/GVIUQgghhMgyVz1cSqkRwJ1AMfCA1vqWuNd/DVxgOWd/oLPWerNSahVQC9QDdVrrY/0JXQghhBAiP6RMuJRSxcA9wOlANfCuUmq21voTcx+t9W3Abcb+o4CrtdabLaf5ltZ6o6+RCyFEmpRSDwEjgQ1a6yOMbR2AJ4BewCrgPK31FuO164FLiPzBeKXW+qUAws5bkydPBmDSpEkBRyLSJW3nPzdDiscDK7TWX2it9wEzgdFJ9h8LzPAjOCGE8NnDwIi4bdcB87TWfYF5xvcopQ4HxgADjGPuNf4AFUKItLkZUuwOrLF8Xw0MtttRKdWayC+mKyybw8DLSqkwcL/WeprHWIUQIiNa69eVUr3iNo8GhhlfPwLMB35jbJ+ptd4LrFRKrSDyB+jCJgm2AEjvSP6StvOfmx6ukM22sMO+o4D/xg0nnqy1HgScCVyulDolzRiFECKbumqtawCMf7sY2+3+2OzexLEJEWPL9j1ZO/e+/fVZO7dw18NVDfS0fN8DWOuw7xjihhO11muNfzcopZ4h8hfi6+mHKoQQTSqdPzaFaBI/mpy9MsI5VcmqhUSm3PRwvQv0VUr1VkqVEUmqZsfvpJQ6EBgKPGfZ1kYp1c78GhgOLPEjcCGE8Ml6pVQ5gPHvBmN7On9sChszZsxgxgwp6c1H0nb+S9nDpbWuU0pdAbxEZFqIh7TWS5VSlxmv32fseg7wstZ6p+XwrsAzSinzWo9rrV/08w0IIUSGZgMXAbcY/z5n2f64UmoqUAH0Bd4JJMI8tXz58qBDEB5J2/nP1TxcWuvngefjtt0X9/3DRJ4Asm77AjgqowiFEMInSqkZRArkOymlqoFJRBKtJ5VSlwCrgR8AGH9YPgl8AtQBl2utpcglDWPGjAk6BOGRtJ3/ZGkfIUSzobV2Wq/kNIf9pwBTshdRYTNGN0QekrbznyztI4QQQgiRZZJwCSGEyIpFixaxaNGioMMQHkjb+U+GFIUQQmTF3LlzAaisrAw4EpEuaTv/ScIlhBAiKwYNGhR0CMIjaTv/ScIlhBDCs3376ykrtV9ictSoURmdO5uzqovkMm07kUgSLiGEEJ6VlRYzauJzqXcUopmTonkhhBBZURraT2lof9BhCA9qa2upra0NOoyCIgmXEEKIrDiq3VKOarc06DCEB1OnTmXq1KlBh1FQZEhRCCFEVuxrkFtMvmrbtm3QIRQc+a9BCCFEVny044igQxAeTZw4MegQCo4MKQohhBBCZJkkXEIIIYQQWSYJlxBCiKzo30bTv40OOgzhwbRp05g2bVrQYRQUqeESQgiRFW2KdwcdgvCopqYm6BAKjiRcQgghsuKTHf2CDkF4NH78+KBDKDiScAkhhMiKXQ2tgw5BeFRRURF0CAVHariEEEIIIbJMEi4hhBBZUdGihooWUguUj+bPn8/8+fODDqOgSMIlhBAiKyparKeixf+3d/9BXtT3HcefcBwqv9QEo/ywghXfFRwlQjGpSSGltpqqmOklBTXa1B/RatJUOknsTMuQTCcmUx2ZBkKA2ppJCnGoRrBEkzHDGaejRYka0bwJBSIniEIVOEC5I9c/dg+/nN+77373vnuf7+739Zhh+H73u/vd12d3b/d9u3uf3R06hqTQ2tpKa2tr6BiFonu4REQkEzvfPT10BElp5syZoSMUjgouERHJxM53x4SOICnNmjUrdITCSVRwmdllwCKgCVjh7nf3+HwW8AiwLR70kLt/Lcm0IiIiIkVXseAysyZgMXAp0AZsMLM17v5yj1F/7u5XpJxWREQKZtjgQ4C6h8ijnTt3AuoeopaS3DQ/A9ji7lvd/QiwCpiT8Pv7M62IiOTY5BGbmTxic+gYksLy5ctZvnx56BiFkuSS4jhgR8n7NuDiMuN91MxeAHYCf+fum6qYVkRECubg0ZNCR5CUxozR/Xe1lqTgGlRmWFeP9xuBs9y93cw+CfwImJRwWhERKaBXDlroCJLSLbfcEjpC4SS5pNgGnFnyfjzRWaxj3H2/u7fHr9cBzWY2Osm0IiIiIkWX5AzXBmCSmU0EXgPmAteUjmBmZwC73b3LzGYQFXJ7gbcrTSsiIiJSdBULLnfvNLM7gMeJuna43903mdmt8edLgRbgNjPrBA4Dc929Cyg7bUZtERGROnLBiJcAeLH9/MBJpFr33HMPAPPnzw+cpDgS9cMVXyZc12PY0pLX3wa+nXRaEREpvqGDO0NHkJTa29tDRygc9TQvIiKZeOHAlNARJKU777wzdITCUcElIiKZ6OhqDh1BUho5cmToCIWT5K8URURERKQfVHCJiEgmzjpxB2eduKPyiFJ31q5dy9q1a0PHKBQVXCIikonThu7ltKF7Q8eQFDZu3MjGjRtDxygU3cMlIiKZ2H54fOgIktIVV1wROkLhqOASEZFM7OkYHTqCpDRt2rTQEQpHlxRFREREMqaCS0REMnHykH2cPGRf6BiSgrvj7qFjFIoKLhERycSkYduYNGxb6BiSwqpVq1i1alXoGIWie7hERCQTb3eMCh1BUjr33HNDRygcFVwiIpKJLYfPDh1BUpo3b17oCIWjS4oiIiIiGdMZLhERwMy2AweAo0Cnu083sw8APwQmANuBz7j7W4EiikiO6QyXiMh7PuHuU919evz+q8AT7j4JeCJ+LwlNH/U800c9HzqGpLBw4UIWLlwYOkahqOASEendHOCB+PUDwNUBs4hIjumSoohIpAv4iZl1Ad9192XA6e6+C8Ddd5nZh4ImzJln908NHUFSWrBgQegIhaOCS0Qkcom774yLqp+a2a9CBxKR4tAlRRERwN13xv+/ATwMzAB2m9kYgPj/N8IlFJE8U8ElIg3PzIab2cju18CfAC8Ba4Ab4tFuAB4JkzCfzjlpK+ectDV0DElh5cqVrFy5MnSMQsnFJcWWlhYAVq9eHTiJiBTU6cDDZgbRfvE/3P0xM9sAPGhmNwKvAp8OmDF3TmneH704HDaHVG/z5s2hIxROLgouEZEsuftW4MIyw/cCswc+UTH8+tDE0BEkpblz54aOUDiJCi4zuwxYBDQBK9z97h6fXwt8JX7bDtzm7i/En22nR2eCNUkuIiJ1bV/nyaEjSErx2V6poYr3cJlZE7AYuByYDMwzs8k9RtsGzHT3C4CvA8t6fN6zM0ERERGRhpHkDNcMYEt8yh0zW0XUGeDL3SO4+3+XjP80ML6WIUVEJH9GN+8BYE/H6MBJpFrPPfccANOmTQucpDiSFFzjgB0l79uAi/sY/0bgxyXvy3UmKCIiBTfhpDZABVcePfroo4AKrlpKUnANKjOsq9yIZvYJooLrYyWD39eZoLs/WX1UERHJkzePfDB0BEnpoosuCh2hcJIUXG3AmSXvxwM7e45kZhcAK4DL47/sAY7vTNDMujsTVMElIlJwv3nnzMojSV268sorQ0conCQdn24AJpnZRDMbCswl6gzwGDP7HeAh4LPuvrlkeG+dCYqIiIg0jIpnuNy908zuAB4n6hbifnffZGa3xp8vBf4R+CCwJP5T0u7uH8p2JphJS0REpK40D+oAoKOrOXASqdaBAwcAGDlyZOAkxZGoHy53Xwes6zFsacnrm4CbykxXtjNBEREpvgtHbgLg2f1TAyeRat17770ALFiwIHCS4lBP8yIikokjv9UhJq9GjBgROkLh6KdBREQy8WL7+aEjSErz588PHaFwktw0LyIiIiL9oIJLREREJGMquEREJBPnDXfOG+6hY0gKy5YtY9kyPRimlnQPl4iIZGJ40+HQESSlXbt2hY5QOCq4REQkEy+3nxs6gqR08803h45QOCq4REQkE4d+Oyx0BElp7NixoSMUju7hEhEREcmYCi4REcnE2BN2MfYE3QuUR+vXr2f9+vWhYxSKCi4REcnE2BN2M/aE3aFjSAqtra20traGjlEouodLREQysfPd00NHkJRmzpwZOkLh5KrgamlpAWD16tWBk4iISCU73x0TOoKkNGvWrNARCkeXFEVERIS39r+T2Xcf6Tia2XfnRa7OcImISH4MG3wIUPcQeXH9wsePva71ult7z5yafE+e6QyXiIhkYvKIzUwesTl0DElB6672dIZLREQycfDoSaEjSEpad7WX24JLN9DXH60TESn1ykELHUFS0rqrPV1SFJGKWlpajhXUIiJSvdye4SqnKGdYitIOyT8VWSIitZH7M1x5+M07DxlrrRHb3BctD2lEF4x4iQtGvBQ6hqSgdVd7hTrDlSWddeo/LUORxjJ0cGfoCJKS1l3tJSq4zOwyYBHQBKxw97t7fD4o/vyTwCHgL919Y5JpQ8pbAdBX3nptS9JcpePlvS0D9T0i9e6FA1NCR5CUtO5qr+IlRTNrAhYDlwOTgXlmNrnHaJcDk+J/twDfqWLaupD0ko8uDfVf6TLU8swfrTNJqqOrmY6u5tAxJAWtu9pLcoZrBrDF3bcCmNkqYA7wcsk4c4DvuXsX8LSZnWJmY4AJCabNrVqfqSh3ECs3j3Jng6S2an02K8t5iFRypOMoQ5ubMvnuLB8HI1IkSQquccCOkvdtwMUJxhmXcNo+fev7z3Lk9EuPG3bXkqf6HHbV7d8FYMqUvk+Jdo9fOu1dS57qc9xy8+ttmnLzSTJeb/NIuhyS2rRpE1B5OSX9HlJkrcV6LM2QtC2Vxi+3zsoNK/2ect9Z63U20JKun2PbQCztNnVw/55U00nfhjY3ceX8RwZ8vmedGO3+f/POmQM+b+kfrbvaS1JwDSozrCvhOEmmrajczjvpsHIHgnIHxtLX5abpa36VDrp9jZd0Hr21r9o2l5tfpVz9zV2LYWkyJF33lbaHnsNKx0+zfiq1L0n+3tqUdJpy8631uq1mu+r26816lEiRnDZ0L6CDdh5p3dVekoKrDShd4uOBnQnHGZpg2j59+brpjB8/vppJjtPSct9x77+x+PPAx1JMU3n8byz+/HGvk4yXxeWkcvn7alO5XOWmzTp3Xypl7Gua0vHKrZ++1ll5fW8/6bz/O5Oux6Trr3RYz3lUs1yTqma76nbVVd/q1zylvmw/nH7fLWFp3dVekoJrAzDJzCYCrwFzgWt6jLMGuCO+R+tiYJ+77zKzNxNM2zBKi5SBLlj6mp/uIWoMSdfzQG0P2u6Kb0/H6NARJCWtu9qrWHC5e6eZ3QE8TtS1w/3uvsnMbo0/XwqsI+oSYgtRtxCf62vaTFrSi+6depqby3VAeL88LZNy6z5P+QeCloeIyMBI1A+Xu68jKqpKhy0ted0F3J502nqng9Dx8r488po/r7nLKVJbJLmTh+wDYF/nyYGTSLW07mpPPc3XgA4mYdRquRd9/YW6lJjm+5YsWcLs2bNrmkPCmTRsGwDP7p8aOIlUS+uu9lRwFVB/LqP2/A6pTyHvBxRJ6u2OUaEjSEpad7WngkuOU68H73rNNdC0HCRPthw+O3QESUnrrvYapuDSgao4tC5lINXz82BFJD8apuBqRCpMGpfWfW2UPA/2UqL+BjeY2Rp3L8TjyUQGylv73+HUUSeGjpFKrR6NpYKrxnSgEymUJM+SrYkiPu9w+qjnAd14nUe1XnfXL3y8Jt8Twtp75tTke+q54GoCeP3110PnEMmNI0eOANDW1hY4STolP+/ZVB7VS/s82FT7r5v+6adVjV/v2ge3A9Bx6P8CJ+m/IUOiw2UR2pJEkdZdfyXdn1baf9VzwTUG4Nprrw2dQyR3CtC1whjgf0OHIP3zYLX/ArYde7UmYIraOPvs6CbybT9rjFv4irTu+mt29eu87P6rnguuDcDHgV3A0cBZRGRgNBHtrDaEDhJL8izZcrT/Emk8fe6/BnV1JfllTUSk8ZjZEGAzMJvoebAbgGsG+hFlIpJ/g0MHEBGpV+7eCXQ/D/YV4EEVWyKShs5wiYiIiGRMZ7hEREREMqaCS0RERCRjKrhEREREMqaCS0RERCRjKrhEREREMla3HZ+a2WXAIqKOxFa4ey669zWzM4HvAWcAvwWWufsiM/sA8ENgArAd+Iy7vxUqZzXiB/g+C7zm7lfktS1mdgqwAjifqLfwvwKcnLXFzP4WuImoDb8EPgcMIwftMLP7gSuAN9z9/HhYr9uTmd0F3EjUeegX3T2/D2SrUqV9oJnNAb5OtJ/pBL7k7k8NeNBAkh4jzOz3gaeBv3D3hnnYbYLtZxbwCO91Kv+Qu39tQEMGlGT7iZfRfUAzsMfdZ/ZnnnV5his+wC8GLgcmA/PMbHLYVIl1AvPd/TzgI8DtcfavAk+4+yTgifh9XvwNUR9E3fLalkXAY+7+e8CFRG3KVVvMbBzwRWB6XLA0AXPJTzv+Hbisx7Cy2eOfm7nAlHiaJfG+ofAS7gOfAC5096lEvzysGNiU4SQ9RsTjfZOoH7WGUcUx9OfuPjX+10jFVsXlE/+CvgS4yt2nAJ/u73zrsuACZgBb3H2rux8BVgG1eVx3xtx9l7tvjF8fIDqojyPK/0A82gPA1WESVsfMxgN/xvE789y1xcxGAX8I/CuAux9x97fJYVuIzkyfFPeCPozoUTO5aIe7Pwn0fBpub9nnAKvc/V133wZsIdo3NIKK+0B3b3f37o4Uh5PsGY9FkfRXoxFaAAADfklEQVQY8QXgP4E3BjJcHcjtMXSAJFk+1xCd9XsVwN37vQ3Va8E1DthR8r4tHpYrZjYB+DDwDHC6u++CqCgDPhQwWjXuA75MdNmiWx7bcjbwJvBvZvYLM1thZsPJWVvc/TXgn4FXiZ7Tt8/df0LO2tFDb9kLsR9IKVHbzexTZvYr4L+IznI1iorLJz4b/Clg6QDmqhdJf3Y+amYvmNmPzWzKwESrC0mWz7nAqWa23syeM7Pr+zvTei24BpUZlqvf3sxsBNFvVl9y9/2h86RhZt332jwXOksNDAEuAr7j7h8GDlK/l916ZWanEv0mNhEYCww3s+vCpspM7vcD/ZCo7e7+cHyJ/Gqi+7kaRZLlcx/wFXdvxIeHJ1k+G4Gz3P1C4F+AH2Weqn4kWT5DgGlEV3j+FPgHMzu3PzOt14KrDTiz5P14ossmuWBmzUTF1g/c/aF48G4zGxN/PoZ8nOK+BLjKzLYTnXL9IzP7PvlsSxvQ5u7PxO9XExVgeWvLHwPb3P1Nd+8AHgL+gPy1o1Rv2XO9H+inqtoeX6r9XTMbnXWwOpFk+UwHVsX7rxaiewDr8lJ7BiouH3ff7+7t8et1QLO2n/eN85i7H3T3PcCTRPf+plavBdcGYJKZTTSzoUQ3zq4JnCkRMxtEdJ/QK+5+b8lHa4Ab4tc3EP11SF1z97vcfby7TyBaBz9z9+vIZ1teB3aYmcWDZgMvk7+2vAp8xMyGxdvabKL7BPPWjlK9ZV8DzDWzE8xsIjAJ+J8A+UKouA80s3PibQAzuwgYCuwd8KRhVFw+7j7R3SfE+6/VwF+7e6OcxUmy/ZxRsv3MIKoHtP285xHg42Y2xMyGARdz/B+PVa0uu4Vw904zu4PoL0uagPvdfVPgWEldAnwW+KWZPR8P+3vgbuBBM7uR6KDZ7794CCivbfkC8IP4B2wrUXcKg8lRW9z9GTNbTXQ5oBP4BbAMGEEO2mFmK4FZwGgzawMW0Mv25O6bzOxBosK4E7i9US4P9bYPNLNb48+XAn8OXG9mHcBhom4PGuKSa8Ll07ASLp8W4DYz6yTafuZq+3lv+bj7K2b2GPAi0T3MK9z9pf7Md1BXV0MsXxEREZFg6vWSooiIiEhhqOASERERyZgKLhEREZGMqeASERERyZgKLhEREZGMqeASERERyZgKLhEREZGMqeASERERydj/A1dTa2cOIqPNAAAAAElFTkSuQmCC", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# plot the posteriors of parameters\n", "m.plot_posteriors()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig__. The mixing of the posterior distribution and autocorrelation looks ok." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Convergence of chains\n", "The Gelman-Rubin statistic is a test of whether the chains in the model converges. The Gelman-Ruben statistic measures the degree of variation between and within chains. Values close to 1 indicate convergence and that there is small variation between chains, i.e. that they end up as the same distribution across chains. A common heuristic is to assume convergence if all values are below 1.1. To run this you need to run multiple models, combine them and perform the Gelman-Rubin statistic:" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# estimate convergence\n", "from kabuki.analyze import gelman_rubin\n", "\n", "models = []\n", "for i in range(3):\n", " m = hddm.HDDMrl(data=data)\n", " m.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")\n", " models.append(m)\n", "\n", "gelman_rubin(models)" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "np.max(list(gelman_rubin(models).values()))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The model seems to have converged, i.e. the Gelman-Rubin statistic is below 1.1 for all parameters. It is important to always run this test, especially for more complex models ([as with separate learning rates for positive and negative prediction errors](#9.-Separate-learning-rates-for-positive-and-negative-prediction-errors)). So now we can combine these three models to get a better approximation of the posterior distribution." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# Combine the models we ran to test for convergence.\n", "m = kabuki.utils.concat_models(models)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Joint posterior distribution\n", "Another test of the model is to look at collinearity. If the estimation of parameters is very codependent (correlation is strong) it can indicate that their variance trades off, in particular if there is a negative correlation. The following plot shows there is generally low correlation across all combinations of parameters. It does not seem to be the case for this dataset, but common for RLDDM is a negative correlation between learning rate and the scaling factor, similar to what's usually observed between learning rate and inverse temperature for RL models that uses softmax as the choice rule (e.g. [Daw, 2011](https://www.oxfordscholarship.com/view/10.1093/acprof:oso/9780199600434.001.0001/acprof-9780199600434-chapter-001))." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "alpha, t, a, v = m.nodes_db.node[[\"alpha\", \"t\", \"a\", \"v\"]]\n", "samples = {\"alpha\": alpha.trace(), \"t\": t.trace(), \"a\": a.trace(), \"v\": v.trace()}\n", "samp = pd.DataFrame(data=samples)\n", "\n", "\n", "def corrfunc(x, y, **kws):\n", " r, _ = stats.pearsonr(x, y)\n", " ax = plt.gca()\n", " ax.annotate(\"r = {:.2f}\".format(r), xy=(0.1, 0.9), xycoords=ax.transAxes)\n", "\n", "\n", "g = sns.PairGrid(samp, palette=[\"red\"])\n", "g.map_upper(plt.scatter, s=10)\n", "g.map_diag(sns.distplot, kde=False)\n", "g.map_lower(sns.kdeplot, cmap=\"Blues_d\")\n", "g.map_lower(corrfunc)\n", "g.savefig(\"matrix_plot.png\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 7. Posterior predictive checks" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "An important test of the model is its ability to recreate the observed data. This can be tested with posterior predictive checks, which involves simulating data using estimated parameters and comparing observed and simulated results." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### extract traces\n", "The first step then is to extract the traces from the estimated model. The function get_traces() gives you the samples (row) from the approaximated posterior distribution for all of the estimated group and subject parameters (column)." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "traces = m.get_traces()\n", "traces.head()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### simulating data\n", "__Now that we have the traces the next step is to simulate data using the estimated parameters. The RLDDM includes a function to simulate data. Here's an example of how to use the simulation-function for RLDDM. This example explains how to generate data with binary outcomes. See [here](#11.-Probabilistic-binary-outcomes-vs.-normally-distributed-outcomes) for an example on simulating data with normally distributed outcomes. Inputs to function:
\n", "a__ = decision threshold
\n", "**t** = non-decision time
\n", "__alpha__ = learning rate
\n", "__pos_alpha__ = defaults to 0. if given it defines the learning rate for positive prediction errors. alpha then becomes the learning rate_ for negative prediction errors.
\n", "__scaler__ = the scaling factor that is multiplied with the difference in q-values to calculate trial-by-trial drift rate
\n", "__p_upper__ = the probability of reward for the option represented by the upper boundary. The current version thus only works for outcomes that are either 1 or 0
\n", "__p_lower__ = the probability of reward for the option represented by the lower boundary.
\n", "__subjs__ = number of subjects to simulate data for.
\n", "__split_by__ = define the condition which makes it easier to append data from different conditions.
\n", "__size__ = number of trials per subject.
" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "hddm.generate.gen_rand_rlddm_data(\n", " a=1,\n", " t=0.3,\n", " alpha=0.2,\n", " scaler=2,\n", " p_upper=0.8,\n", " p_lower=0.2,\n", " subjs=1,\n", " split_by=0,\n", " size=10,\n", ")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__How to interpret columns in the resulting dataframe__
\n", "__q_up__ = expected reward for option represented by upper boundary
\n", "__q_low__ = expected reward for option represented by lower boundary
\n", "__sim_drift__ = the drift rate for each trial calculated as: (q_up-q_low)*scaler
\n", "__response__ = simulated choice
\n", "__rt__ = simulated response time
\n", "__feedback__ = observed feedback for chosen option
\n", "__subj_idx__ = subject id (starts at 0)
\n", "__split_by__ = condition as integer
\n", "__trial__ = current trial (starts at 1)
" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Simulate data with estimated parameter values and compare to observed data\n", "Now that we know how to extract traces and simulate data we can combine this to create a dataset similar to our observed data. This process is currently not automated but the following is an example code using the dataset we analyzed above." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "from tqdm import tqdm # progress tracker\n", "\n", "# create empty dataframe to store simulated data\n", "sim_data = pd.DataFrame()\n", "# create a column samp to be used to identify the simulated data sets\n", "data[\"samp\"] = 0\n", "# load traces\n", "traces = m.get_traces()\n", "# decide how many times to repeat simulation process. repeating this multiple times is generally recommended,\n", "# as it better captures the uncertainty in the posterior distribution, but will also take some time\n", "for i in tqdm(range(1, 51)):\n", " # randomly select a row in the traces to use for extracting parameter values\n", " sample = np.random.randint(0, traces.shape[0] - 1)\n", " # loop through all subjects in observed data\n", " for s in data.subj_idx.unique():\n", " # get number of trials for each condition.\n", " size0 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 0)].trial.unique()\n", " )\n", " size1 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 1)].trial.unique()\n", " )\n", " size2 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 2)].trial.unique()\n", " )\n", " # set parameter values for simulation\n", " a = traces.loc[sample, \"a_subj.\" + str(s)]\n", " t = traces.loc[sample, \"t_subj.\" + str(s)]\n", " scaler = traces.loc[sample, \"v_subj.\" + str(s)]\n", " alphaInv = traces.loc[sample, \"alpha_subj.\" + str(s)]\n", " # take inverse logit of estimated alpha\n", " alpha = np.exp(alphaInv) / (1 + np.exp(alphaInv))\n", " # simulate data for each condition changing only values of size, p_upper, p_lower and split_by between conditions.\n", " sim_data0 = hddm.generate.gen_rand_rlddm_data(\n", " a=a,\n", " t=t,\n", " scaler=scaler,\n", " alpha=alpha,\n", " size=size0,\n", " p_upper=0.8,\n", " p_lower=0.2,\n", " split_by=0,\n", " )\n", " sim_data1 = hddm.generate.gen_rand_rlddm_data(\n", " a=a,\n", " t=t,\n", " scaler=scaler,\n", " alpha=alpha,\n", " size=size1,\n", " p_upper=0.7,\n", " p_lower=0.3,\n", " split_by=1,\n", " )\n", " sim_data2 = hddm.generate.gen_rand_rlddm_data(\n", " a=a,\n", " t=t,\n", " scaler=scaler,\n", " alpha=alpha,\n", " size=size2,\n", " p_upper=0.6,\n", " p_lower=0.4,\n", " split_by=2,\n", " )\n", " # append the conditions\n", " sim_data0 = sim_data0.append([sim_data1, sim_data2], ignore_index=True)\n", " # assign subj_idx\n", " sim_data0[\"subj_idx\"] = s\n", " # identify that these are simulated data\n", " sim_data0[\"type\"] = \"simulated\"\n", " # identify the simulated data\n", " sim_data0[\"samp\"] = i\n", " # append data from each subject\n", " sim_data = sim_data.append(sim_data0, ignore_index=True)\n", "# combine observed and simulated data\n", "ppc_data = data[\n", " [\"subj_idx\", \"response\", \"split_by\", \"rt\", \"trial\", \"feedback\", \"samp\"]\n", "].copy()\n", "ppc_data[\"type\"] = \"observed\"\n", "ppc_sdata = sim_data[\n", " [\"subj_idx\", \"response\", \"split_by\", \"rt\", \"trial\", \"feedback\", \"type\", \"samp\"]\n", "].copy()\n", "ppc_data = ppc_data.append(ppc_sdata)\n", "ppc_data.to_csv(\"ppc_data_tutorial.csv\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plotting\n", "Now that we have a dataframe with both observed and simulated data we can plot to see whether the simulated data are able to capture observed choice and reaction times. To capture the uncertainty in the simulated data we want to identify how much choice and reaction differs across the simulated data sets. A good measure of this is to calculate the highest posterior density/highest density interval for summary scores of the generated data. Below we calculate highest posterior density with an alpha set to 0.1, which means that we are describing the range of the 90% most likely values." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# for practical reasons we only look at the first 40 trials for each subject in a given condition\n", "plot_ppc_data = ppc_data[ppc_data.trial < 41].copy()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Choice" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# bin trials to for smoother estimate of response proportion across learning\n", "plot_ppc_data[\"bin_trial\"] = pd.cut(\n", " plot_ppc_data.trial, 11, labels=np.linspace(0, 10, 11)\n", ").astype(\"int64\")\n", "# calculate means for each sample\n", "sums = (\n", " plot_ppc_data.groupby([\"bin_trial\", \"split_by\", \"samp\", \"type\"])\n", " .mean()\n", " .reset_index()\n", ")\n", "# calculate the overall mean response across samples\n", "ppc_sim = sums.groupby([\"bin_trial\", \"split_by\", \"type\"]).mean().reset_index()\n", "# initiate columns that will have the upper and lower bound of the hpd\n", "ppc_sim[\"upper_hpd\"] = 0\n", "ppc_sim[\"lower_hpd\"] = 0\n", "for i in range(0, ppc_sim.shape[0]):\n", " # calculate the hpd/hdi of the predicted mean responses across bin_trials\n", " hdi = pymc.utils.hpd(\n", " sums.response[\n", " (sums[\"bin_trial\"] == ppc_sim.bin_trial[i])\n", " & (sums[\"split_by\"] == ppc_sim.split_by[i])\n", " & (sums[\"type\"] == ppc_sim.type[i])\n", " ],\n", " alpha=0.1,\n", " )\n", " ppc_sim.loc[i, \"upper_hpd\"] = hdi[1]\n", " ppc_sim.loc[i, \"lower_hpd\"] = hdi[0]\n", "# calculate error term as the distance from upper bound to mean\n", "ppc_sim[\"up_err\"] = ppc_sim[\"upper_hpd\"] - ppc_sim[\"response\"]\n", "ppc_sim[\"low_err\"] = ppc_sim[\"response\"] - ppc_sim[\"lower_hpd\"]\n", "ppc_sim[\"model\"] = \"RLDDM_single_learning\"\n", "ppc_sim.to_csv(\"ppc_choicedata_tutorial.csv\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# plotting evolution of choice proportion for best option across learning for observed and simulated data.\n", "fig, axs = plt.subplots(figsize=(15, 5), nrows=1, ncols=3, sharex=True, sharey=True)\n", "for i in range(0, 3):\n", " ax = axs[i]\n", " d = ppc_sim[(ppc_sim.split_by == i) & (ppc_sim.type == \"simulated\")]\n", " ax.errorbar(\n", " d.bin_trial,\n", " d.response,\n", " yerr=[d.low_err, d.up_err],\n", " label=\"simulated\",\n", " color=\"orange\",\n", " )\n", " d = ppc_sim[(ppc_sim.split_by == i) & (ppc_sim.type == \"observed\")]\n", " ax.plot(d.bin_trial, d.response, linewidth=3, label=\"observed\")\n", " ax.set_title(\"split_by = %i\" % i, fontsize=20)\n", " ax.set_ylabel(\"mean response\")\n", " ax.set_xlabel(\"trial\")\n", "plt.legend()\n", "fig.savefig(\"PPCchoice.pdf\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The plots display the rate of choosing the best option (response = 1) across learning and condition. The model generates data (orange) that closely follows the observed behavior (blue), with the exception of overpredicting performance early in the most difficult condition (split_by=2). Uncertainty in the generated data is captured by the 90% highest density interval of the means across simulated datasets." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### RT" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# set reaction time to be negative for lower bound responses (response=0)\n", "plot_ppc_data[\"reaction time\"] = np.where(\n", " plot_ppc_data[\"response\"] == 1, plot_ppc_data.rt, 0 - plot_ppc_data.rt\n", ")\n", "# plotting evolution of choice proportion for best option across learning for observed and simulated data. We use bins of trials because plotting individual trials would be very noisy.\n", "g = sns.FacetGrid(plot_ppc_data, col=\"split_by\", hue=\"type\")\n", "g.map(sns.kdeplot, \"reaction time\", bw=0.05).set_ylabels(\"Density\")\n", "g.add_legend()\n", "g.savefig(\"PPCrt_dist.pdf\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ Density plots of observed and predicted reaction time across conditions. RTs for lower boundary choices (i.e. worst option choices) are set to be negative (0-RT) to be able to separate upper and lower bound responses." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 8. Parameter recovery\n", "To validate the RLDDM we ran a parameter recovery study to test to which degree the model can recover the parameter values used to simulate data. To do this we generated 81 synthetic datasets with 50 subjects performing 70 trials each. The 81 datasets were simulated using all combinations of three plausible parameter values for decision threshold, non-decision time, learning rate and the scaling parameter onto drift rate." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Estimated values split by simulated vales \n", "We can plot simulated together with the estimated values to test the models ability to recover parameters, and to see if there are any values that are more difficult to recover than others." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "param_recovery = hddm.load_csv(\"recovery_sim_est_rlddm.csv\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "g = sns.catplot(x=\"a\", y=\"e_a\", data=param_recovery, palette=\"Set1\")\n", "g.set_axis_labels(\"Simulated threshold\", \"Estimated threshold\")\n", "plt.title(\"Decision threshold\")\n", "g.savefig(\"Threshold_recovery.pdf\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "g = sns.catplot(x=\"alpha\", y=\"e_alphaT\", data=param_recovery, palette=\"Set1\")\n", "g.set_axis_labels(\"Simulated alpha\", \"Estimated alpha\")\n", "plt.title(\"Learning rate\")\n", "g.savefig(\"Alpha_recovery.pdf\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "g = sns.catplot(x=\"scaler\", y=\"e_v\", data=param_recovery, palette=\"Set1\")\n", "g.set_axis_labels(\"Simulated scaling\", \"Estimated scaling\")\n", "plt.title(\"Scaling drift rate\")\n", "g.savefig(\"Scaler_recovery.pdf\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "g = sns.catplot(x=\"t\", y=\"e_t\", data=param_recovery, palette=\"Set1\")\n", "g.set_axis_labels(\"Simulated NDT\", \"Estimated NDT\")\n", "plt.title(\"Non-decision time\")\n", "g.savefig(\"NDT_recovery.pdf\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The correlation between simulated and estimated parameter values are high, which means recovery is good. There is somewhat worse recovery for the learning rate and scaling parameter, which makes sense given that they to a degree can explain the same variance (see below)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 9. Separate learning rates for positive and negative prediction errors\n", "Several studies have reported differences in updating of expected rewards following positive and negative prediction errors (e.g. to capture differences between D1 and D2 receptor function). To model asymmetric updating rates for positive and negative prediction errors you can set dual=True in the model. This will produce two estimated learning rates; alpha and pos_alpha, of which alpha then becomes the estimated learning rate for negative prediction errors." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# set dual=True to model separate learning rates for positive and negative prediction errors.\n", "m_dual = hddm.HDDMrl(data, dual=True)\n", "# set sample and burn-in\n", "m_dual.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")\n", "# print stats to get an overview of posterior distribution of estimated parameters\n", "m_dual.print_stats()" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "m_dual.plot_posteriors()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ There's more autocorrelation in this model compared to the one with a single learning rate. First, let's test whether it converges." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# estimate convergence\n", "models = []\n", "for i in range(3):\n", " m = hddm.HDDMrl(data=data, dual=True)\n", " m.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")\n", " models.append(m)\n", "\n", "# get max gelman-statistic value. shouldn't be higher than 1.1\n", "np.max(list(gelman_rubin(models).values()))" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "gelman_rubin(models)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Convergence looks good, i.e. no parameters with gelman-rubin statistic > 1.1." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# Create a new model that has all traces concatenated\n", "# of individual models.\n", "m_dual = kabuki.utils.concat_models(models)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "And then we can have a look at the joint posterior distribution:" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "alpha, t, a, v, pos_alpha = m_dual.nodes_db.node[[\"alpha\", \"t\", \"a\", \"v\", \"pos_alpha\"]]\n", "samples = {\n", " \"alpha\": alpha.trace(),\n", " \"pos_alpha\": pos_alpha.trace(),\n", " \"t\": t.trace(),\n", " \"a\": a.trace(),\n", " \"v\": v.trace(),\n", "}\n", "samp = pd.DataFrame(data=samples)\n", "\n", "\n", "def corrfunc(x, y, **kws):\n", " r, _ = stats.pearsonr(x, y)\n", " ax = plt.gca()\n", " ax.annotate(\"r = {:.2f}\".format(r), xy=(0.1, 0.9), xycoords=ax.transAxes)\n", "\n", "\n", "g = sns.PairGrid(samp, palette=[\"red\"])\n", "g.map_upper(plt.scatter, s=10)\n", "g.map_diag(sns.distplot, kde=False)\n", "g.map_lower(sns.kdeplot, cmap=\"Blues_d\")\n", "g.map_lower(corrfunc)\n", "g.savefig(\"matrix_plot.png\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The correlation between parameters is generally low. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Posterior predictive check\n", "The DIC for this dual learning rate model is better than for the single learning rate model. We can therefore check whether we can detect this improvement in the ability to recreate choice and RT patterns:" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# create empty dataframe to store simulated data\n", "sim_data = pd.DataFrame()\n", "# create a column samp to be used to identify the simulated data sets\n", "data[\"samp\"] = 0\n", "# get traces, note here we extract traces from m_dual\n", "traces = m_dual.get_traces()\n", "# decide how many times to repeat simulation process. repeating this multiple times is generally recommended as it better captures the uncertainty in the posterior distribution, but will also take some time\n", "for i in tqdm(range(1, 51)):\n", " # randomly select a row in the traces to use for extracting parameter values\n", " sample = np.random.randint(0, traces.shape[0] - 1)\n", " # loop through all subjects in observed data\n", " for s in data.subj_idx.unique():\n", " # get number of trials for each condition.\n", " size0 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 0)].trial.unique()\n", " )\n", " size1 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 1)].trial.unique()\n", " )\n", " size2 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 2)].trial.unique()\n", " )\n", " # set parameter values for simulation\n", " a = traces.loc[sample, \"a_subj.\" + str(s)]\n", " t = traces.loc[sample, \"t_subj.\" + str(s)]\n", " scaler = traces.loc[sample, \"v_subj.\" + str(s)]\n", " # when generating data with two learning rates pos_alpha represents learning rate for positive prediction errors and alpha for negative prediction errors\n", " alphaInv = traces.loc[sample, \"alpha_subj.\" + str(s)]\n", " pos_alphaInv = traces.loc[sample, \"pos_alpha_subj.\" + str(s)]\n", " # take inverse logit of estimated alpha and pos_alpha\n", " alpha = np.exp(alphaInv) / (1 + np.exp(alphaInv))\n", " pos_alpha = np.exp(pos_alphaInv) / (1 + np.exp(pos_alphaInv))\n", " # simulate data for each condition changing only values of size, p_upper, p_lower and split_by between conditions.\n", " sim_data0 = hddm.generate.gen_rand_rlddm_data(\n", " a=a,\n", " t=t,\n", " scaler=scaler,\n", " alpha=alpha,\n", " pos_alpha=pos_alpha,\n", " size=size0,\n", " p_upper=0.8,\n", " p_lower=0.2,\n", " split_by=0,\n", " )\n", " sim_data1 = hddm.generate.gen_rand_rlddm_data(\n", " a=a,\n", " t=t,\n", " scaler=scaler,\n", " alpha=alpha,\n", " pos_alpha=pos_alpha,\n", " size=size1,\n", " p_upper=0.7,\n", " p_lower=0.3,\n", " split_by=1,\n", " )\n", " sim_data2 = hddm.generate.gen_rand_rlddm_data(\n", " a=a,\n", " t=t,\n", " scaler=scaler,\n", " alpha=alpha,\n", " pos_alpha=pos_alpha,\n", " size=size2,\n", " p_upper=0.6,\n", " p_lower=0.4,\n", " split_by=2,\n", " )\n", " # append the conditions\n", " sim_data0 = sim_data0.append([sim_data1, sim_data2], ignore_index=True)\n", " # assign subj_idx\n", " sim_data0[\"subj_idx\"] = s\n", " # identify that these are simulated data\n", " sim_data0[\"type\"] = \"simulated\"\n", " # identify the simulated data\n", " sim_data0[\"samp\"] = i\n", " # append data from each subject\n", " sim_data = sim_data.append(sim_data0, ignore_index=True)\n", "# combine observed and simulated data\n", "ppc_dual_data = data[\n", " [\"subj_idx\", \"response\", \"split_by\", \"rt\", \"trial\", \"feedback\", \"samp\"]\n", "].copy()\n", "ppc_dual_data[\"type\"] = \"observed\"\n", "ppc_dual_sdata = sim_data[\n", " [\"subj_idx\", \"response\", \"split_by\", \"rt\", \"trial\", \"feedback\", \"type\", \"samp\"]\n", "].copy()\n", "ppc_dual_data = ppc_dual_data.append(ppc_dual_sdata)" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# for practical reasons we only look at the first 40 trials for each subject in a given condition\n", "plot_ppc_dual_data = ppc_dual_data[ppc_dual_data.trial < 41].copy()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Choice" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# bin trials to for smoother estimate of response proportion across learning\n", "plot_ppc_dual_data[\"bin_trial\"] = pd.cut(\n", " plot_ppc_dual_data.trial, 11, labels=np.linspace(0, 10, 11)\n", ").astype(\"int64\")\n", "# calculate means for each sample\n", "sums = (\n", " plot_ppc_dual_data.groupby([\"bin_trial\", \"split_by\", \"samp\", \"type\"])\n", " .mean()\n", " .reset_index()\n", ")\n", "# calculate the overall mean response across samples\n", "ppc_dual_sim = sums.groupby([\"bin_trial\", \"split_by\", \"type\"]).mean().reset_index()\n", "# initiate columns that will have the upper and lower bound of the hpd\n", "ppc_dual_sim[\"upper_hpd\"] = 0\n", "ppc_dual_sim[\"lower_hpd\"] = 0\n", "for i in range(0, ppc_dual_sim.shape[0]):\n", " # calculate the hpd/hdi of the predicted mean responses across bin_trials\n", " hdi = pymc.utils.hpd(\n", " sums.response[\n", " (sums[\"bin_trial\"] == ppc_dual_sim.bin_trial[i])\n", " & (sums[\"split_by\"] == ppc_dual_sim.split_by[i])\n", " & (sums[\"type\"] == ppc_dual_sim.type[i])\n", " ],\n", " alpha=0.1,\n", " )\n", " ppc_dual_sim.loc[i, \"upper_hpd\"] = hdi[1]\n", " ppc_dual_sim.loc[i, \"lower_hpd\"] = hdi[0]\n", "# calculate error term as the distance from upper bound to mean\n", "ppc_dual_sim[\"up_err\"] = ppc_dual_sim[\"upper_hpd\"] - ppc_dual_sim[\"response\"]\n", "ppc_dual_sim[\"low_err\"] = ppc_dual_sim[\"response\"] - ppc_dual_sim[\"lower_hpd\"]\n", "ppc_dual_sim[\"model\"] = \"RLDDM_dual_learning\"" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# plotting evolution of choice proportion for best option across learning for observed and simulated data.\n", "fig, axs = plt.subplots(figsize=(15, 5), nrows=1, ncols=3, sharex=True, sharey=True)\n", "for i in range(0, 3):\n", " ax = axs[i]\n", " d = ppc_dual_sim[(ppc_dual_sim.split_by == i) & (ppc_dual_sim.type == \"simulated\")]\n", " ax.errorbar(\n", " d.bin_trial,\n", " d.response,\n", " yerr=[d.low_err, d.up_err],\n", " label=\"simulated\",\n", " color=\"orange\",\n", " )\n", " d = ppc_sim[(ppc_dual_sim.split_by == i) & (ppc_dual_sim.type == \"observed\")]\n", " ax.plot(d.bin_trial, d.response, linewidth=3, label=\"observed\")\n", " ax.set_title(\"split_by = %i\" % i, fontsize=20)\n", " ax.set_ylabel(\"mean response\")\n", " ax.set_xlabel(\"trial\")\n", "plt.legend()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The plots display the rate of choosing the best option (response = 1) across learning and condition. The model generates data (orange) that closely follows the observed behavior (blue), with the exception of performance early in the most difficult condition (split_by=2)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### PPC for single vs. dual learning rate\n", "To get a better sense of differences in ability to predict data between the single and dual learning rate model we can plot them together:" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# plotting evolution of choice proportion for best option across learning for observed and simulated data. Compared for model with single and dual learning rate.\n", "fig, axs = plt.subplots(figsize=(15, 5), nrows=1, ncols=3, sharex=True, sharey=True)\n", "for i in range(0, 3):\n", " ax = axs[i]\n", " d_single = ppc_sim[(ppc_sim.split_by == i) & (ppc_sim.type == \"simulated\")]\n", " # slightly move bin_trial to avoid overlap in errorbars\n", " d_single[\"bin_trial\"] += 0.2\n", " ax.errorbar(\n", " d_single.bin_trial,\n", " d_single.response,\n", " yerr=[d_single.low_err, d_single.up_err],\n", " label=\"simulated_single\",\n", " color=\"orange\",\n", " )\n", " d_dual = ppc_dual_sim[\n", " (ppc_dual_sim.split_by == i) & (ppc_dual_sim.type == \"simulated\")\n", " ]\n", " ax.errorbar(\n", " d_dual.bin_trial,\n", " d_dual.response,\n", " yerr=[d_dual.low_err, d_dual.up_err],\n", " label=\"simulated_dual\",\n", " color=\"green\",\n", " )\n", " d = ppc_sim[(ppc_dual_sim.split_by == i) & (ppc_dual_sim.type == \"observed\")]\n", " ax.plot(d.bin_trial, d.response, linewidth=3, label=\"observed\")\n", " ax.set_title(\"split_by = %i\" % i, fontsize=20)\n", " ax.set_ylabel(\"mean response\")\n", " ax.set_xlabel(\"trial\")\n", "plt.xlim(-0.5, 10.5)\n", "plt.legend()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The predictions from the model with two learning rates are not very different from the model with single learning rate, and a similar overprediction of performance early on for the most difficult condition (split_by =2)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### RT" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "plot_ppc_data[\"type_compare\"] = np.where(\n", " plot_ppc_data[\"type\"] == \"observed\",\n", " plot_ppc_data[\"type\"],\n", " \"simulated_single_learning\",\n", ")\n", "plot_ppc_dual_data[\"type_compare\"] = np.where(\n", " plot_ppc_dual_data[\"type\"] == \"observed\",\n", " plot_ppc_dual_data[\"type\"],\n", " \"simulated_dual_learning\",\n", ")\n", "dual_vs_single_pcc = plot_ppc_data.append(plot_ppc_dual_data)\n", "dual_vs_single_pcc[\"reaction time\"] = np.where(\n", " dual_vs_single_pcc[\"response\"] == 1,\n", " dual_vs_single_pcc.rt,\n", " 0 - dual_vs_single_pcc.rt,\n", ")\n", "# plotting evolution of choice proportion for best option across learning for observed and simulated data. We use bins of trials because plotting individual trials would be very noisy.\n", "g = sns.FacetGrid(dual_vs_single_pcc, col=\"split_by\", hue=\"type_compare\", height=5)\n", "g.map(sns.kdeplot, \"reaction time\", bw=0.01).set_ylabels(\"Density\")\n", "g.add_legend()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ Again there's not a big difference between the two models. Both models slightly overpredict performance for the medium (split_by =1) and hard (split_by = 2) conditions, as identified by lower densities for the negative (worst option choices) in the simulated compared to observed data." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Transform alpha and pos_alpha\n", "To interpret the parameter estimates for alpha and pos_alpha you have to transform them with the inverse logit where learning rate for negative prediction error is alpha and learning rate for positive prediction errors is pos_alpha. For this dataset the learning rate is estimated to be higher for positive than negative prediction errors." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# plot alpha for positive and negative learning rate\n", "traces = m_dual.get_traces()\n", "neg_alpha = np.exp(traces[\"alpha\"]) / (1 + np.exp(traces[\"alpha\"]))\n", "pos_alpha = np.exp(traces[\"pos_alpha\"]) / (1 + np.exp(traces[\"pos_alpha\"]))\n", "sns.kdeplot(\n", " neg_alpha, color=\"r\", label=\"neg_alpha: \" + str(np.round(np.mean(neg_alpha), 3))\n", ")\n", "sns.kdeplot(\n", " pos_alpha, color=\"b\", label=\"pos_alpha: \" + str(np.round(np.mean(pos_alpha), 3))\n", ")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The positive learning rate is estimated to be stronger than the negative learning rate. Sticky choice, tendencies to repeat choices, could be driving some of this difference. The current model does not allow to test for this, however, but it could be tested in the future if we implement a regression version of RLDDM (similar to HDDMRegressor)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Simulate data with learning rates for positive and negative prediction errors\n", "Here's how you would simulate data with a learning rate for positive and negative predictions of 0.2 and 0.4, respectively:" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "hddm.generate.gen_rand_rlddm_data(\n", " a=1, t=0.3, alpha=0.2, pos_alpha=0.4, scaler=2, p_upper=0.8, p_lower=0.2, size=10\n", ")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 10. depends_on vs. split_by\n", "HDDMrl can be used to estimate separate parameters just as in the standard HDDM. But in RL you typically estimate the same learning rates and inverse temperature across conditions. That's one reason why you have to specify condition in the split_by-column instead of depends_on. (The other is that if you use depends_on the expected rewards will not get updated properly). But depends_on is still useful, for example if you want to estimate the effect of group on parameters. As an example we can simulate a dataset with two groups that have different decision thresholds:" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "data1 = hddm.generate.gen_rand_rlddm_data(\n", " a=1, t=0.3, alpha=0.4, scaler=2, p_upper=0.8, p_lower=0.2, subjs=50, size=50\n", ")\n", "data1[\"group\"] = \"group1\"\n", "data2 = hddm.generate.gen_rand_rlddm_data(\n", " a=2, t=0.3, alpha=0.4, scaler=2, p_upper=0.8, p_lower=0.2, subjs=50, size=50\n", ")\n", "data2[\"group\"] = \"group2\"\n", "group_data = data1.append(data2)\n", "group_data[\"q_init\"] = 0.5\n", "m = hddm.HDDMrl(\n", " group_data, depends_on={\"v\": \"group\", \"a\": \"group\", \"t\": \"group\", \"alpha\": \"group\"}\n", ")\n", "m.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# the plot shows that the model was able to recover the different decision threshold across groups.\n", "a_group1, a_group2 = m.nodes_db.node[[\"a(group1)\", \"a(group2)\"]]\n", "hddm.analyze.plot_posterior_nodes([a_group1, a_group2])\n", "plt.xlabel(\"decision threshold\")\n", "plt.ylabel(\"Posterior probability\")\n", "plt.xlim(0.7, 2.3)\n", "plt.title(\"Posterior of decision threshold group means\")" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 11. Probabilistic binary outcomes vs. normally distributed outcomes\n", "The examples so far have all been using a task structure where the outcomes are binary and probabilistic. But the model can also be applied to other types of outcomes. Here we show how you can generate and model data with normally distributed outcomes. As you will see you don't have to do any modifications to the model estimation process, but you have to change the input for generating data. Also note that the scaling parameter (v) will scale negatively with the values of the observed outcomes because the combined drift rate needs to be plausible." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# This is how we generated data so far, defining the probability of reward (1) for actions/stimuli associated with upper and lower boundary.\n", "# binary probabilistic outcomes\n", "hddm.generate.gen_rand_rlddm_data(\n", " a=2, t=0.3, scaler=2, alpha=0.2, size=10, p_upper=0.2, p_lower=0.8\n", ")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# If instead the outcomes are drawn from a normal distribution you will have to set binary_outcome to False and instead of p_upper and p_upper define the mean (mu) and sd\n", "# of the normal distribution for both alternatives. Note that we change the initial q-value to 0, and that we reduce the scaling factor.\n", "# normally distributed outcomes\n", "hddm.generate.gen_rand_rlddm_data(\n", " a=2,\n", " t=0.3,\n", " scaler=0.2,\n", " alpha=0.2,\n", " size=10,\n", " mu_upper=8,\n", " mu_lower=2,\n", " sd_upper=1,\n", " sd_lower=1,\n", " binary_outcome=False,\n", " q_init=0,\n", ")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# We can generate a dataset where 30 subjects perform 50 trials each. Note that we set the scaler to be lower than for the binary outcomes as otherwise\n", "# the resulting drift will be unrealistically high.\n", "norm_data = hddm.generate.gen_rand_rlddm_data(\n", " a=2,\n", " t=0.3,\n", " scaler=0.2,\n", " alpha=0.2,\n", " size=50,\n", " subjs=30,\n", " mu_upper=8,\n", " mu_lower=2,\n", " sd_upper=2,\n", " sd_lower=2,\n", " binary_outcome=False,\n", " q_init=0,\n", ")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# and then we can do estimation as usual\n", "# but first we need to define inital q-value\n", "norm_data[\"q_init\"] = 0\n", "m_norm = hddm.HDDMrl(norm_data)\n", "m_norm.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")\n", "m_norm.print_stats()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 12. HDDMrlRegressor\n", "\n", "As of version 0.7.6. HDDM includes a module for estimating the impact of continuous regressor onto RLDDM parameters. The module, called HDDMrlRegressor, works the same way as the HDDMRegressor for the normal DDM. The method allows estimation of the association of e.g. neural measures onto parameters. To illustrate the method we extend the function to generate rlddm_data by adding a normally distributed regressor and including a coefficient called 'neural'.Note that to run the HDDMrlRegressor you need to include alpha when specifying the model. For more information on how to set up regressor models look at the tutorial for HDDM." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# function to generate rlddm-data that adds a neural regressor to decision threshold\n", "def gen_rand_reg_rlddm_data(\n", " a,\n", " t,\n", " scaler,\n", " alpha,\n", " neural,\n", " size=1,\n", " p_upper=1,\n", " p_lower=0,\n", " z=0.5,\n", " q_init=0.5,\n", " split_by=0,\n", " subjs=1,\n", "):\n", " all_data = []\n", " n = size\n", " # set sd for variables to generate subject-parameters from group distribution\n", " sd_t = 0.02\n", " sd_a = 0.1\n", " sd_alpha = 0.1\n", " sd_v = 0.25\n", " # save parameter values as group-values\n", " tg = t\n", " ag = a\n", " alphag = alpha\n", " scalerg = scaler\n", " for s in range(0, subjs):\n", " t = (\n", " np.maximum(0.05, np.random.normal(loc=tg, scale=sd_t, size=1))\n", " if subjs > 1\n", " else tg\n", " )\n", " a = (\n", " np.maximum(0.05, np.random.normal(loc=ag, scale=sd_a, size=1))\n", " if subjs > 1\n", " else ag\n", " )\n", " alpha = (\n", " np.minimum(\n", " np.minimum(\n", " np.maximum(0.001, np.random.normal(loc=alphag, scale=sd_a, size=1)),\n", " alphag + alphag,\n", " ),\n", " 1,\n", " )\n", " if subjs > 1\n", " else alphag\n", " )\n", " scaler = (\n", " np.random.normal(loc=scalerg, scale=sd_v, size=1) if subjs > 1 else scalerg\n", " )\n", " # create a normalized regressor that is combined with the neural coefficient to create trial-by-trial values for decision threshold\n", " neural_reg = np.random.normal(0, 1, size=n)\n", " q_up = np.tile([q_init], n)\n", " q_low = np.tile([q_init], n)\n", " response = np.tile([0.5], n)\n", " feedback = np.tile([0.5], n)\n", " rt = np.tile([0], n)\n", " rew_up = np.random.binomial(1, p_upper, n).astype(float)\n", " rew_low = np.random.binomial(1, p_lower, n).astype(float)\n", " sim_drift = np.tile([0], n)\n", " subj_idx = np.tile([s], n)\n", " d = {\n", " \"q_up\": q_up,\n", " \"q_low\": q_low,\n", " \"sim_drift\": sim_drift,\n", " \"rew_up\": rew_up,\n", " \"rew_low\": rew_low,\n", " \"response\": response,\n", " \"rt\": rt,\n", " \"feedback\": feedback,\n", " \"subj_idx\": subj_idx,\n", " \"split_by\": split_by,\n", " \"trial\": 1,\n", " \"neural_reg\": neural_reg,\n", " }\n", " df = pd.DataFrame(data=d)\n", " df = df[\n", " [\n", " \"q_up\",\n", " \"q_low\",\n", " \"sim_drift\",\n", " \"rew_up\",\n", " \"rew_low\",\n", " \"response\",\n", " \"rt\",\n", " \"feedback\",\n", " \"subj_idx\",\n", " \"split_by\",\n", " \"trial\",\n", " \"neural_reg\",\n", " ]\n", " ]\n", " # generate data trial-by-trial using the Intercept (a), regressor (neural_reg) and coefficient (neural) for decision threshold.\n", " data, params = hddm.generate.gen_rand_data(\n", " {\n", " \"a\": a + neural * df.loc[0, \"neural_reg\"],\n", " \"t\": t,\n", " \"v\": df.loc[0, \"sim_drift\"],\n", " \"z\": z,\n", " },\n", " subjs=1,\n", " size=1,\n", " )\n", " df.loc[0, \"response\"] = data.response[0]\n", " df.loc[0, \"rt\"] = data.rt[0]\n", " if data.response[0] == 1.0:\n", " df.loc[0, \"feedback\"] = df.loc[0, \"rew_up\"]\n", " else:\n", " df.loc[0, \"feedback\"] = df.loc[0, \"rew_low\"]\n", "\n", " for i in range(1, n):\n", " df.loc[i, \"trial\"] = i + 1\n", " df.loc[i, \"q_up\"] = (\n", " df.loc[i - 1, \"q_up\"] * (1 - df.loc[i - 1, \"response\"])\n", " ) + (\n", " (df.loc[i - 1, \"response\"])\n", " * (\n", " df.loc[i - 1, \"q_up\"]\n", " + (alpha * (df.loc[i - 1, \"rew_up\"] - df.loc[i - 1, \"q_up\"]))\n", " )\n", " )\n", " df.loc[i, \"q_low\"] = (\n", " df.loc[i - 1, \"q_low\"] * (df.loc[i - 1, \"response\"])\n", " ) + (\n", " (1 - df.loc[i - 1, \"response\"])\n", " * (\n", " df.loc[i - 1, \"q_low\"]\n", " + (alpha * (df.loc[i - 1, \"rew_low\"] - df.loc[i - 1, \"q_low\"]))\n", " )\n", " )\n", " df.loc[i, \"sim_drift\"] = (df.loc[i, \"q_up\"] - df.loc[i, \"q_low\"]) * (scaler)\n", " data, params = hddm.generate.gen_rand_data(\n", " {\n", " \"a\": a + neural * df.loc[i, \"neural_reg\"],\n", " \"t\": t,\n", " \"v\": df.loc[i, \"sim_drift\"],\n", " \"z\": z,\n", " },\n", " subjs=1,\n", " size=1,\n", " )\n", " df.loc[i, \"response\"] = data.response[0]\n", " df.loc[i, \"rt\"] = data.rt[0]\n", " if data.response[0] == 1.0:\n", " df.loc[i, \"feedback\"] = df.loc[i, \"rew_up\"]\n", " else:\n", " df.loc[i, \"feedback\"] = df.loc[i, \"rew_low\"]\n", " all_data.append(df)\n", " all_data = pd.concat(all_data, axis=0)\n", " all_data = all_data[\n", " [\n", " \"q_up\",\n", " \"q_low\",\n", " \"sim_drift\",\n", " \"response\",\n", " \"rt\",\n", " \"feedback\",\n", " \"subj_idx\",\n", " \"split_by\",\n", " \"trial\",\n", " \"neural_reg\",\n", " ]\n", " ]\n", "\n", " return all_data" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# Create data with function defined above.\n", "# This will create trial-by-trial values for decision threshold (a) by adding the coefficient neural (here set to 0.2)\n", "# multiplied by a normalized regressor (neural_reg) to the 'Intercept' value of a (here set to 1)\n", "data_neural = gen_rand_reg_rlddm_data(\n", " a=1,\n", " t=0.3,\n", " scaler=2,\n", " alpha=0.2,\n", " neural=0.2,\n", " size=100,\n", " p_upper=0.7,\n", " p_lower=0.3,\n", " subjs=25,\n", ")\n", "data_neural[\"q_init\"] = 0.5\n", "data_neural.head()" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# run a regressor model estimating the impact of 'neural' on decision threshold a. This should estimate the coefficient a_neural_reg to be 0.2\n", "# to run the HDDMrlRegressor you need to include alpha\n", "m_reg = hddm.HDDMrlRegressor(\n", " data_neural, \"a ~ neural_reg\", include=[\"v\", \"a\", \"t\", \"alpha\"]\n", ")\n", "m_reg.sample(1000, burn=250)\n", "m_reg.print_stats()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 13. Regular RL without RT\n", "HDDMrl also includes a module to run an RL-model that uses softmax to transform q-values to probability of choosing options associated with upper (response=1) or lower (response=0) boundary. To run this model you type hddm.Hrl instead of hddm.HDDMrl. The setup is the same as for HDDMrl, and for now, the model won't run if you don't include an rt-column. This will be fixed for a future version, but for now, if you don't have RTs you can just create an rt-column where you set all rts to e.g. 0.5. You can choose to estimate separate learning rates for positive and negative learning rate by setting dual=True (see [here](#9.-Separate-learning-rates-for-positive-and-negative-prediction-errors) for more information). The model will by default estimate posterior distributions for the alpha and v parameters. The probability of choosing upper boundary is captured as:

\n", "      $p_{up} =(e^{-2*z*d_t}-1)/ (e^{-2*d_t}-1)$,

\n", "where ${d_t}=q_{up_t}-q_{low}*v$ and z represents starting point (which for now is fixed to be 0.5).
\n", "This calculation is equivalent to soft-max transformation when z=0.5." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# run the model by calling hddm.Hrl (instead of hddm.HDDM for normal model and hddm.HDDMrl for rlddm-model)\n", "m_rl = hddm.Hrl(data)\n", "# set sample and burn-in\n", "m_rl.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")\n", "# print stats to get an overview of posterior distribution of estimated parameters\n", "m_rl.print_stats()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Parameter estimates from the pure RL-model are a bit different compared to the RLDDM. This is to be expected as probability of choice in DDM is dependent both on the decsision threshold and the scaled difference in q-values, whereas the RL model only uses the scaled difference in q-values. " ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "m_rl.plot_posteriors()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ Mixing and autocorrelation looks good." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# estimate convergence\n", "models = []\n", "for i in range(3):\n", " m = hddm.Hrl(data=data)\n", " m.sample(1500, burn=500, dbname=\"traces.db\", db=\"pickle\")\n", " models.append(m)\n", "# get max gelman-statistic value. shouldn't be higher than 1.1\n", "np.max(list(gelman_rubin(models).values()))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Convergence looks good, i.e. no parameters with gelman-rubin statistic > 1.1." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# Create a new model that has all traces concatenated\n", "# of individual models.\n", "m_rl = kabuki.utils.concat_models(models)" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "alpha, v = m_rl.nodes_db.node[[\"alpha\", \"v\"]]\n", "samples = {\"alpha\": alpha.trace(), \"v\": v.trace()}\n", "samp = pd.DataFrame(data=samples)\n", "\n", "\n", "def corrfunc(x, y, **kws):\n", " r, _ = stats.pearsonr(x, y)\n", " ax = plt.gca()\n", " ax.annotate(\"r = {:.2f}\".format(r), xy=(0.1, 0.9), xycoords=ax.transAxes)\n", "\n", "\n", "g = sns.PairGrid(samp, palette=[\"red\"])\n", "g.map_upper(plt.scatter, s=10)\n", "g.map_diag(sns.distplot, kde=False)\n", "g.map_lower(sns.kdeplot, cmap=\"Blues_d\")\n", "\n", "g.map_lower(corrfunc)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The correlation in the posterior distribution for alpha and v/scaling is somewhat negative." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Posterior predictive check\n", "We can also do posterior predictive check on the RL-model by generating new data with hddm.generate.gen_rand_rl_data." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# create empty dataframe to store simulated data\n", "sim_data = pd.DataFrame()\n", "# create a column samp to be used to identify the simulated data sets\n", "data[\"samp\"] = 0\n", "# load traces\n", "traces = m_rl.get_traces()\n", "# decide how many times to repeat simulation process. repeating this multiple times is generally recommended as it better captures the uncertainty in the posterior distribution, but will also take some time\n", "for i in tqdm(range(1, 51)):\n", " # randomly select a row in the traces to use for extracting parameter values\n", " sample = np.random.randint(0, traces.shape[0] - 1)\n", " # loop through all subjects in observed data\n", " for s in data.subj_idx.unique():\n", " # get number of trials for each condition.\n", " size0 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 0)].trial.unique()\n", " )\n", " size1 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 1)].trial.unique()\n", " )\n", " size2 = len(\n", " data[(data[\"subj_idx\"] == s) & (data[\"split_by\"] == 2)].trial.unique()\n", " )\n", " # set parameter values for simulation\n", " scaler = traces.loc[sample, \"v_subj.\" + str(s)]\n", " alphaInv = traces.loc[sample, \"alpha_subj.\" + str(s)]\n", " # take inverse logit of estimated alpha\n", " alpha = np.exp(alphaInv) / (1 + np.exp(alphaInv))\n", " # simulate data for each condition changing only values of size, p_upper, p_lower and split_by between conditions.\n", " sim_data0 = hddm.generate.gen_rand_rl_data(\n", " scaler=scaler, alpha=alpha, size=size0, p_upper=0.8, p_lower=0.2, split_by=0\n", " )\n", " sim_data1 = hddm.generate.gen_rand_rl_data(\n", " scaler=scaler, alpha=alpha, size=size1, p_upper=0.7, p_lower=0.3, split_by=1\n", " )\n", " sim_data2 = hddm.generate.gen_rand_rl_data(\n", " scaler=scaler, alpha=alpha, size=size2, p_upper=0.6, p_lower=0.4, split_by=2\n", " )\n", " # append the conditions\n", " sim_data0 = sim_data0.append([sim_data1, sim_data2], ignore_index=True)\n", " # assign subj_idx\n", " sim_data0[\"subj_idx\"] = s\n", " # identify that these are simulated data\n", " sim_data0[\"type\"] = \"simulated\"\n", " # identify the simulated data\n", " sim_data0[\"samp\"] = i\n", " # append data from each subject\n", " sim_data = sim_data.append(sim_data0, ignore_index=True)\n", "# combine observed and simulated data\n", "ppc_rl_data = data[\n", " [\"subj_idx\", \"response\", \"split_by\", \"trial\", \"feedback\", \"samp\"]\n", "].copy()\n", "ppc_rl_data[\"type\"] = \"observed\"\n", "ppc_rl_sdata = sim_data[\n", " [\"subj_idx\", \"response\", \"split_by\", \"trial\", \"feedback\", \"type\", \"samp\"]\n", "].copy()\n", "ppc_rl_data = ppc_rl_data.append(ppc_rl_sdata)" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# for practical reasons we only look at the first 40 trials for each subject in a given condition\n", "plot_ppc_rl_data = ppc_rl_data[ppc_rl_data.trial < 41].copy()" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# bin trials to for smoother estimate of response proportion across learning\n", "plot_ppc_rl_data[\"bin_trial\"] = pd.cut(\n", " plot_ppc_rl_data.trial, 11, labels=np.linspace(0, 10, 11)\n", ").astype(\"int64\")\n", "# calculate means for each sample\n", "sums = (\n", " plot_ppc_rl_data.groupby([\"bin_trial\", \"split_by\", \"samp\", \"type\"])\n", " .mean()\n", " .reset_index()\n", ")\n", "# calculate the overall mean response across samples\n", "ppc_rl_sim = sums.groupby([\"bin_trial\", \"split_by\", \"type\"]).mean().reset_index()\n", "# initiate columns that will have the upper and lower bound of the hpd\n", "ppc_rl_sim[\"upper_hpd\"] = 0\n", "ppc_rl_sim[\"lower_hpd\"] = 0\n", "for i in range(0, ppc_rl_sim.shape[0]):\n", " # calculate the hpd/hdi of the predicted mean responses across bin_trials\n", " hdi = pymc.utils.hpd(\n", " sums.response[\n", " (sums[\"bin_trial\"] == ppc_rl_sim.bin_trial[i])\n", " & (sums[\"split_by\"] == ppc_rl_sim.split_by[i])\n", " & (sums[\"type\"] == ppc_rl_sim.type[i])\n", " ],\n", " alpha=0.1,\n", " )\n", " ppc_rl_sim.loc[i, \"upper_hpd\"] = hdi[1]\n", " ppc_rl_sim.loc[i, \"lower_hpd\"] = hdi[0]\n", "# calculate error term as the distance from upper bound to mean\n", "ppc_rl_sim[\"up_err\"] = ppc_rl_sim[\"upper_hpd\"] - ppc_rl_sim[\"response\"]\n", "ppc_rl_sim[\"low_err\"] = ppc_rl_sim[\"response\"] - ppc_rl_sim[\"lower_hpd\"]\n", "ppc_rl_sim[\"model\"] = \"RL\"" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# plotting evolution of choice proportion for best option across learning for observed and simulated data. Compared for RL and RLDDM models, both with single learnign rate.\n", "fig, axs = plt.subplots(figsize=(15, 5), nrows=1, ncols=3, sharex=True, sharey=True)\n", "for i in range(0, 3):\n", " ax = axs[i]\n", " d_single = ppc_sim[(ppc_sim.split_by == i) & (ppc_sim.type == \"simulated\")]\n", " # slightly move bin_trial to avoid overlap in errorbars\n", " d_single[\"bin_trial\"] += 0.2\n", " ax.errorbar(\n", " d_single.bin_trial,\n", " d_single.response,\n", " yerr=[d_single.low_err, d_single.up_err],\n", " label=\"simulated_RLDDM\",\n", " color=\"orange\",\n", " )\n", " ax = axs[i]\n", " d_rl = ppc_rl_sim[(ppc_rl_sim.split_by == i) & (ppc_rl_sim.type == \"simulated\")]\n", " ax.errorbar(\n", " d_rl.bin_trial,\n", " d_rl.response,\n", " yerr=[d_rl.low_err, d_rl.up_err],\n", " label=\"simulated_RL\",\n", " color=\"green\",\n", " )\n", " ax = axs[i]\n", " d = ppc_sim[(ppc_dual_sim.split_by == i) & (ppc_dual_sim.type == \"observed\")]\n", " ax.plot(d.bin_trial, d.response, linewidth=3, label=\"observed\")\n", " ax.set_title(\"split_by = %i\" % i, fontsize=20)\n", " ax.set_ylabel(\"mean response\")\n", " ax.set_xlabel(\"trial\")\n", "plt.xlim(-0.5, 10.5)\n", "plt.legend()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Fig.__ The predicted choice for the RL-model is very similar to what was predicted in the RLDDM. That is not surprising given that they use the same calculation to get the choice likelihood. The difference between them is instead that the RLDDM could potentially detect the unique contribution of the scaling/drift parameter and the decision threshold onto choice." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Misprediction across learning\n", "Another way to visualize this is to look at how the predicted choice misses on the observed across learning, i.e. predicted-observed. As for the other plots we see that the two methods are very similar. " ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "# rl\n", "error_prediction = (\n", " plot_ppc_rl_data.groupby([\"split_by\", \"type\", \"bin_trial\"])[\"response\"]\n", " .mean()\n", " .reset_index()\n", ")\n", "ep = error_prediction.pivot_table(\n", " index=[\"split_by\", \"bin_trial\"], columns=\"type\", values=\"response\"\n", ").reset_index()\n", "ep[\"diff\"] = ep[\"simulated\"] - ep[\"observed\"]\n", "ep[\"model\"] = \"RL\"\n", "# rlddm\n", "error_prediction = (\n", " plot_ppc_data.groupby([\"split_by\", \"type\", \"bin_trial\"])[\"response\"]\n", " .mean()\n", " .reset_index()\n", ")\n", "ep_rlddm = error_prediction.pivot_table(\n", " index=[\"split_by\", \"bin_trial\"], columns=\"type\", values=\"response\"\n", ").reset_index()\n", "ep_rlddm[\"diff\"] = ep_rlddm[\"simulated\"] - ep_rlddm[\"observed\"]\n", "ep_rlddm[\"model\"] = \"RLDDM\"\n", "# combine\n", "ep = ep.append(ep_rlddm)\n", "# plot\n", "g = sns.relplot(\n", " x=\"bin_trial\",\n", " y=\"diff\",\n", " col=\"split_by\",\n", " hue=\"model\",\n", " kind=\"line\",\n", " ci=False,\n", " data=ep,\n", " palette=\"Set2_r\",\n", ")\n", "g.map(plt.axhline, y=0, ls=\":\", c=\".5\")" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "interpreter": { "hash": "a445e8b86844784ad439c05402e9f190af217ba9ac3e10cbacead74ef6ea1c6a" }, "kernelspec": { "display_name": "Python 3.6.13 64-bit ('hddm': conda)", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.6.13" } }, "nbformat": 4, "nbformat_minor": 4 }