{ "metadata": { "name": "", "signature": "sha256:a83952d6db125e3ad95b33ed5997004e5195d461b6b307ca746acb757fb9e050" }, "nbformat": 3, "nbformat_minor": 0, "worksheets": [ { "cells": [ { "cell_type": "code", "collapsed": false, "input": [ "%matplotlib inline\n", "import matplotlib.pyplot as plt\n", "\n", "# Some nice default configuration for plots\n", "plt.rcParams['figure.figsize'] = 10, 7.5\n", "plt.rcParams['axes.grid'] = True\n", "plt.gray()\n", "\n", "import numpy as np\n", "import pandas as pd" ], "language": "python", "metadata": {}, "outputs": [ { "metadata": {}, "output_type": "display_data", "text": [ "" ] } ], "prompt_number": 1 }, { "cell_type": "heading", "level": 1, "metadata": {}, "source": [ "5. Measuring Performance in Regression Models" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "For models predicting a numeric outcome, some measure of accuracy is typically used to evaluate the effectiveness of the model. However, there are different ways to measure accuracy, each with its own nuance. To understand the strengths and weaknesses of a particular model, relying solely on a single metric is problematic. Visualizations of the model fit, particularly residual plots, are critical to understand whether the model is fit for purpose." ] }, { "cell_type": "heading", "level": 2, "metadata": {}, "source": [ "5.1 Quantitative Measures of Performance" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "When the outcome is a number, the most common method for characterizing a model's predictive capabilities is to use the root mean squared error (RMSE). This metric is a function of the model residuals, which are the observed values minus the model predictions. The value is usually interpreted as either how far (on average) the residuals are from zero or as the average distance between the observed values and the model predictions.\n", "\n", "Another common metric is the coefficient of determination, i.e., $R^2$. This value can be interpreted as the proportion of the information in the data that is explained by the model, i.e., the proportion of variation is explained by the model. While it is an easily interpretable statistic, the practitioner must remember that $R^2$ is a measure of correlation, not accuracy. Also, it is dependent on the variation in the outcome. Practically, this dependence on the outcome variance can also have a drastic effect on how the model is viewed.\n", "\n", "In some cases, the goal of the model is to simply rank new samples, where the *rank correlation* between the observed and predicted values might be a more appropriate metric. The rank correlation takes the ranks of the observed outcome values and evaluates how close these are to ranks of the model predictions. To calculate this value, the ranks of the observed and predicted outcomes are obtained and the correlation coefficient between these ranks is calculated, which is known as Spearman's rank correlation." ] }, { "cell_type": "heading", "level": 2, "metadata": {}, "source": [ "5.2 The Variance-Bias Trade-off" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The MSE can be decomposed into more specific pieces. Formally, the MSE of a model is $$\\text{MSE} = {1\\over n} \\sum_{i=1}^n (y_i - \\hat{y}_i)^2,$$ where $y_i$ is the outcome and $\\hat{y}_i$ is the model prediction of that sample's outcome. If we assume that the data points are statistically independent and that the residuals have a theoretical mean of zero and a constant variance of $\\sigma^2$, then $$E[\\text{MSE}] = \\sigma^2 + (\\text{Model Bias})^2 + \\text{Model Variance}.$$ The first part $(\\sigma^2)$ is usually called \"irreducible noise\" and cannot be eliminated by modeling. The second term is the squared bias of the model. This reflects how close the functional form of the model can get to the true relationship between the predictors and the outcome. The last term is the model variance." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "An extreme example of models that are either high bias or high variance." ] }, { "cell_type": "code", "collapsed": false, "input": [ "# a simulated sin wave\n", "X = np.random.uniform(2, 10, 100)\n", "y = np.sin(X) + np.random.normal(0, 0.2, 100)" ], "language": "python", "metadata": {}, "outputs": [], "prompt_number": 2 }, { "cell_type": "code", "collapsed": false, "input": [ "# high bias estimate\n", "est_bias = np.zeros(100)\n", "est_bias[:50] = np.mean(y[np.argsort(X)[:50]])\n", "est_bias[50:] = np.mean(y[np.argsort(X)[50:]])\n", "\n", "# high variance estimate\n", "def movingaverage(values, window):\n", " '''calculate simple moving average'''\n", " weigths = np.repeat(1.0, window)/window\n", " #including valid will REQUIRE there to be enough datapoints.\n", " #for example, if you take out valid, it will start @ point one,\n", " #not having any prior points, so itll be 1+0+0 = 1 /3 = .3333\n", " smas = np.convolve(values, weigths, 'valid')\n", " return smas\n", "\n", "est_var = movingaverage(y[np.argsort(X)], 3) # MA(3)" ], "language": "python", "metadata": {}, "outputs": [], "prompt_number": 3 }, { "cell_type": "code", "collapsed": false, "input": [ "plt.scatter(X, y)\n", "\n", "# plot high bias estimate\n", "plt_bias, = plt.plot(np.insert(X[np.argsort(X)], 50, np.nan), \n", " np.insert(est_bias, 50, np.nan), # insert discontinuous point\n", " color='g', linewidth=2)\n", "\n", "# plot high variance estimate\n", "plt_var, = plt.plot(X[np.argsort(X)][2:], est_var, color='r')\n", "\n", "plt.xlabel(\"Predictor\")\n", "plt.ylabel(\"Outcome\")\n", "plt.legend([plt_bias, plt_var], ['High bias model', 'High variance model'])" ], "language": "python", "metadata": {}, "outputs": [ { "metadata": {}, "output_type": "pyout", "prompt_number": 4, "text": [ "" ] }, { "metadata": {}, "output_type": "display_data", "png": "iVBORw0KGgoAAAANSUhEUgAAAmUAAAHSCAYAAACtqLx3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkQj4fysMgtvtkkH9AAAAAElFTkSuQmCC\n", "text": [ "" ] } ], "prompt_number": 4 }, { "cell_type": "markdown", "metadata": {}, "source": [ "It is generally true that more complex models can have very high variance, which leads to over-fitting. On the other hand, simple models tend not to over-fit, but under-fit if they are not flexible enough to model the true relationship (thus high bias). Also, highly correlated predictors can lead to collinearity issues and this can greatly increase the model variance. Increase the bias in the model to greatly reduce the model variance is a way to mitigate the problem of collinearity. This is referred to as the *variance-bias trade-off*." ] } ], "metadata": {} } ] }