{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# An Introduction to Fisher Forecasting\n",
    "\n",
    "*Zack Li, 7/27/2018*\n",
    "\n",
    "As a cosmologist, you'll probably face the following question.\n",
    "\n",
    "> Will this dataset be able to measure what I'm interested in?\n",
    "\n",
    "More concrete examples from my own research include\n",
    "* Will Planck lensing be useful for measuring properties of dark matter? \n",
    "* Should I run some N-body simulations which would take millions of CPU hours, in order to measure a parameter?\n",
    "* Will the hierarchy of neutrino masses be tested with LSST galaxy lensing?\n",
    "* Having performed a Monte Carlo analysis for a $\\Lambda CDM$ extension using the latest CMB data, are my results plausible?\n",
    "\n",
    "If your observables have **Gaussian uncertainties**, Fisher information matrices can help answer these questions. They perform the change of variables which turns uncertainties of correlated observables (i.e. the CMB power spectrum) into uncertainties of correlated parameters (i.e. the inputs for the $\\Lambda CDM$ model). It is a crude (but useful!) error propagation, and applies to many cosmological data sets like CMB, weak lensing, and BAO.\n",
    "\n",
    "![Image of ACTPol](../images/angelapano.jpg)\n",
    "\n",
    "Since Fisher forecasting is a numerical technique, this guide is packaged as an interactive notebook. It's intended for advanced undergraduates and early graduate students, and the goal is to build intuition and be able to compute a Fisher matrix by the end. We'll cover the basic methods for computing Fisher information matrices, using the power spectrum of the cosmic microwave background (CMB) as an example. Useful references are [Coe 2009](https://arxiv.org/abs/0906.4123) for implementation tips, as well as the more comprehensive [Verde 2009](https://arxiv.org/abs/0911.3105) which I have drawn heavily from.\n",
    "\n",
    "To use this notebook, you'll need to know a little calculus and cosmology, be able to state Bayes theorem, and have [CLASS](http://class-code.net/) installed with the Python wrapper."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline \n",
    "%config InlineBackend.figure_format = 'retina' # I use a Mac\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from classy import Class  # CLASS python wrapper"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## A Change of Variables\n",
    "\n",
    "Suppose you have some probability distribution $P(x)$ in terms of the variable $x$, and want to transform it so it's in terms of the variable $y$. What is  $Q(y)$, the transformed distribution? Changing variables should conserve the probability of a set of events, so in terms of differentials,\n",
    "\n",
    "$$P(x) \\, dx = Q(y) \\, dy.$$\n",
    "\n",
    "Dividing, we derive the familiar chain rule relating $P(x)$ and $Q(y)$,\n",
    "\n",
    "$$P(x) = Q(y) \\, \\left|\\frac{dy}{dx}\\right|.$$\n",
    "\n",
    "That is, you can transform probability distributions into new variables by using derivatives. In the case of multivariate Gaussian measurement uncertainty, the probability distributions can be described in terms of a covariance matrix. This is one way to interpret Fisher information matrix methods: you are **transforming the covariance matrix** describing your observables (i.e. measurement uncertainties) into the covariance matrix describing your parameters (i.e. uncertainties on numbers like $H_0$). \n",
    "\n",
    "What do you need to get there? Following the example above, you need to know the *covariance matrix* of your observable, and you need to know the *derivatives* between the observables and parameters. \n",
    "\n",
    "We now define the Fisher information matrix, $\\mathbf{F}$, with elements $F_{ij}$. \n",
    "\n",
    "$$F_{ij} = - \\left\\langle\\frac{\\partial^2}{\\partial \\theta_i \\theta_j} \\mathrm{ln}\\,f\\right\\rangle.$$\n",
    "\n",
    "Here $f$ is the likelihood, and $\\theta_i$ and $\\theta_j$ are parameters you are interested in. Each element $F_{ij}$ of the Fisher information matrix is defined as the Hessian of the log likelihood, averaged over realizations of your data. This is a more geometric interpretion of the Fisher information matrix: it describes the **curvature of the log likelihood**. If the likelihood is \"sharp\" or \"peaked\", that means our experiment is very good at measuring the parameters. With such a peaked likelihood, the curvature of the log likelihood at the peak will be high. Similarly, if the likelihood is wide and spread out, the curvature will be low, as will the Fisher information.\n",
    "\n",
    "![Sketch of curvature](../images/curvature_sketch.png)\n",
    "\n",
    "The Fisher matrix is useful because of the *Cramér-Rao Bound*, which states that the variance of an unbiased estimator is at least as high as the inverse of the Fisher information matrix. For parameters $x$ and $y$, the covariance $\\sigma_{xy}$ is bounded with the inequality\n",
    "\n",
    "$$\\sigma_{xy} \\geq F_{xy}^{-1}.$$\n",
    "\n",
    "In other words, the inverse of the Fisher information matrix gives you best possible covariance matrix for the parameters you are trying to estimate. A nice informal derivation of this fact is on the [Wikipedia page](https://en.wikipedia.org/wiki/Fisher_information) for Fisher information.\n",
    "\n",
    "We now return to the change of variables interpretation. Recall that the probabilities must sum to 1, so that $\\int f\\, d^n\\theta = 1$. Taking a derivative of this identity and combining with the definition of the Fisher information matrix, one can derive the transformation rule for a set of variables $\\{\\lambda_i\\}$ to the set $\\{\\theta_i\\}$ for the Fisher matrix (more details in [Tegmark 1996](https://arxiv.org/abs/astro-ph/9611174)),\n",
    "\n",
    "$$ \\mathbf{F}^{\\theta} = \\mathbf{J}^T \\mathbf{F}^{\\theta} \\mathbf{J},$$\n",
    "\n",
    "The matrix $\\mathbf{J}$ is the Jacobian, with elements $J_{ij} \\equiv \\frac{\\partial \\lambda_i}{\\partial \\theta_j}$.\n",
    "\n",
    "## Our Example: The CMB Power Spectrum\n",
    "\n",
    "We've been speaking generally about covariances and derivatives. We'll now make a brief detour and discuss the main example of this notebook, the angular power spectrum $C_{\\ell}$ of the fluctuations in the cosmic microwave background (CMB). There are many details here, and I'd recommend a textbook like *Modern Cosmology* by Dodelson if you want to learn more.\n",
    "\n",
    "A short summary which might stir some memories if you've taken a cosmology course: \n",
    "\n",
    "* The CMB has fluctuations on the order of $10^{-4}$ in temperature and polarization, and the statistical properties of the fluctuations depend on cosmological parameters like the cold dark matter density $\\Omega_c$, the baryon density $\\Omega_b$, and the amplitude of the primordial perturbations $A_s$.\n",
    "* Since the CMB is on the (spherical) sky, these fluctuations can be decomposed into the spherical harmonics, which form an orthonormal and complete basis. The decomposition yields coefficients $a_{\\ell m}$. \n",
    "* The CMB map is a Gaussian random field (GRF), which for our purposes means the $a_{\\ell m}$ are independent and Gaussian-distributed over different realizations of the universe. We can characterize the statistical properties of the CMB with a variance $C_{\\ell}$, by averaging the spherical harmonic coefficients over noise realizations of the CMB,\n",
    "$$\\langle a_{\\ell m} a_{\\ell' m'} \\rangle = \\delta_{\\ell \\ell'} \\delta_{m m'} C_{\\ell}.$$\n",
    "\n",
    "We don't need a very detailed knowledge of this to make progress here; if you want, just think of $C_{\\ell}$ just as a set of numbers you can measure, indexed by $\\ell$. The Boltzmann code CLASS can generate theory curves of $C_{\\ell}$ given a set of cosmological parameters. Let's generate the $\\Lambda CDM$ temperature (TT) spectrum now with CLASS. We'll name it `fiducial` because we'll use it later as the fiducial power spectrum for Fisher forecasts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define the CLASS input dictionary, use defaults\n",
    "params = {\n",
    "    'output': 'tCl lCl',\n",
    "    'l_max_scalars': 2000,\n",
    "    'lensing': 'yes',\n",
    "    'omega_cdm': 0.120,\n",
    "    'omega_b': 0.0224, \n",
    "    'h': 0.674\n",
    "}\n",
    "\n",
    "# The usual CLASS code for computing C_ell\n",
    "cosmo = Class()\n",
    "cosmo.set(params)\n",
    "cosmo.compute()\n",
    "fiducial = cosmo.lensed_cl(2000)\n",
    "cosmo.struct_cleanup()\n",
    "cosmo.empty()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We are choosing a parametrization $\\{ \\omega_{cdm}, \\omega_b, h\\}$ with $\\omega_{cdm} = \\Omega_{c} h^2$, $\\omega_{b} = \\Omega_{b} h^2$, $h = H_0 / (100\\:\\mathrm{km/s/Mpc})$. Let's plot it in a common way to view power spectra, $\\ell^2 C_{\\ell}^{TT}$. The $\\ell^2$ factor is so you can see the high-$\\ell$ features, which are hard to see otherwise due to Silk damping."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 313,
       "width": 423
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(fiducial['ell'], fiducial['ell']**2*fiducial['tt'])\n",
    "plt.xlabel(r'$\\ell$')\n",
    "plt.ylabel(r'$\\ell^2 C_{\\ell}^{TT}$');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compute a Derivative\n",
    "We've suggested that derivatives are important for a change of variables. For the CMB, we can actually write the Fisher matrix for the parameters in terms of the Fisher matrix for $C_{\\ell}$, which we'll describe later. We can write the transformation rule as\n",
    "\n",
    "$$ \\mathbf{F}^{\\theta} = \\mathbf{J}^T \\mathbf{F}^{C_{\\ell}} \\mathbf{J}.$$\n",
    "\n",
    "The variable we have is $C_{\\ell}$, and the variables we want are a set of cosmological parameters $\\{ \\theta_i \\}$, for example the normalized Hubble constant $h$. Thus, we will be interested in computing a Jacobian $\\mathbf{J}$ which corresponds to \n",
    "\n",
    "$$ \\frac{\\partial C_{\\ell}^{XY}}{\\partial \\theta_i} $$\n",
    "\n",
    "for observable $XY \\in \\{ TT, TE, EE \\}$. As an example, we will now estimate $\\partial C_{\\ell}^{TT} / \\partial h$, the derivative of the TT power spectrum with respect to the normalized Hubble constant. This is just a matter of choosing a step size $\\Delta h$, and then computing $\\Delta C_{\\ell} / \\Delta h$.\n",
    "\n",
    "Good step sizes for Fisher typically are $\\sim 1\\%$ of the parameter itself. Smaller steps are usually better for the derivative, but going too small can lead to numerical instability."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def utility_function_call_CLASS(input_dict, l_max=2000):\n",
    "    \"\"\"Compute Cl with this utility function, repeat less code.\"\"\"\n",
    "    cosmo = Class()\n",
    "    cosmo.set(input_dict)\n",
    "    cosmo.compute()\n",
    "    temp_cl = cosmo.lensed_cl(l_max)\n",
    "    cosmo.struct_cleanup()\n",
    "    cosmo.empty()\n",
    "    return temp_cl\n",
    "\n",
    "\n",
    "# Define the CLASS input dictionary, use defaults\n",
    "params = {\n",
    "    'output': 'tCl lCl',\n",
    "    'l_max_scalars': 2000,\n",
    "    'lensing': 'yes',\n",
    "    'omega_cdm': 0.120,\n",
    "    'omega_b': 0.0224, \n",
    "    'h': 0.674\n",
    "}\n",
    "\n",
    "# for left and right sides of the derivative, copy\n",
    "# the dict above, and then change the parameter\n",
    "# we are interested in. \n",
    "h_step = 0.01\n",
    "left_params = params.copy()\n",
    "left_params['h'] = params['h'] - h_step\n",
    "right_params = params.copy()\n",
    "right_params['h'] = params['h'] + h_step\n",
    "\n",
    "# get the C_l^TT and then compute the derivative!\n",
    "cl_tt_left = utility_function_call_CLASS(left_params)['tt']\n",
    "cl_tt_right = utility_function_call_CLASS(right_params)['tt']\n",
    "dCltt_dh = (cl_tt_right - cl_tt_left) / (2 * h_step)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's plot the derivative! We'll plot it but normalize by $C_{\\ell}^{TT}$ so it's more clear. The warning message below is just because the CLASS output has zeroes for $\\ell < 3$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/zequnl/anaconda3/lib/python3.6/site-packages/ipykernel/__main__.py:1: RuntimeWarning: invalid value encountered in true_divide\n",
      "  if __name__ == '__main__':\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 303,
       "width": 439
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot( dCltt_dh / fiducial['tt'] )\n",
    "plt.ylabel(r'$(\\partial C_{\\ell}^{TT} / \\partial h) / C_{\\ell}^{TT}$')\n",
    "plt.xlabel(r'$\\ell$');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "-----\n",
    "## Assignment 1\n",
    "Write a function `get_deriv()` which takes in a parameter name (like `omega_b`), a channel (like `'tt'`), and a stepsize (like `0.01`) and returns the derivative $\\partial C_{\\ell}^{XY} / \\partial \\theta_i$. Test it out by reproducing the plot above."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# fill me in!\n",
    "def get_deriv():\n",
    "    pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "-----"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Nice Gaussian Facts\n",
    "\n",
    "We now have to revisit the angle brackets in the definition of the Fisher information matrix. Here is the definition again,\n",
    "\n",
    "$$F_{ij} = - \\left\\langle\\frac{\\partial^2}{\\partial \\theta_i \\theta_j} \\mathrm{ln}\\,f\\right\\rangle.$$\n",
    "\n",
    "The $\\langle \\cdots \\rangle$ refer to an average over *data realizations*, or in other words, an average of each possible way that the data could have turned out given the assumed noise properties. For continuous variables, this requires taking an integral in data space that is weighted by the probability density of obtaining that data. In the case of CMB data, the power spectrum can have order 50 bins, thus requiring a 50-dimensional integral. Cancel.\n",
    "\n",
    "If the measurement uncertainty is described by a multivariate Gaussian, then we can avoid doing this (see [Tegmark et al. 1996](https://arxiv.org/abs/astro-ph/9603021) for the derivation in the CMB context, an identical derivation exists for lensing). The signal in the sky for temperature (T) as well as curl-free (E) and divergence-free (B) polarization (i.e. E and B modes) then consists of a set over $m$, \n",
    "\n",
    "$$ (a_{\\ell}^T, a_{\\ell}^E, a_{\\ell}^B)$$\n",
    "\n",
    "We don't have multiple skies, but a GRF has the nice property that we can estimate the variance on $a_{\\ell m}$ by looking at different parts of the sky. Unfortunately this estimator is limited by the sample size for a specific angle scale on the sky, and we have only $2 \\ell + 1$ samples. There is uncertainty on the estimate of $C_{\\ell}$, we have $(\\Delta C_{\\ell})^2  = \\frac{2}{2\\ell+1} C_{\\ell}^2$. This is the so-called *cosmic variance*.\n",
    "\n",
    "We therefore have for each positive integer $\\ell$ an observable $C_{\\ell}$ that we can measure. The observable $C_{\\ell}$ has intrinsic variance $(\\Delta C_{\\ell})^2 = \\frac{2}{2\\ell+1} C_{\\ell}^2$. Assuming that the B-modes are negligible :( we can write for correlated T and E spectra a matrix proportional to the covariance for each $\\ell$,\n",
    "\n",
    "$$ \\mathbf{C}_{\\ell} \\equiv \n",
    "  \\left( {\\begin{array}{cc}\n",
    "   C_{\\ell}^{TT} + N_{\\ell}^{TT} & C_{\\ell}^{TE} \\\\\n",
    "   C_{\\ell}^{TE} & C_{\\ell}^{EE} + N_{\\ell}^{EE} \\\\\n",
    "  \\end{array} } \\right)\n",
    " $$\n",
    "\n",
    "where $N_{\\ell}^{TT}$ and $N_{\\ell}^{EE}$ refer to additional instrumental noise in the temperature and polarization, which are added to the noise from cosmic variance. The cross-spectrum noise $N_{\\ell}^{TE}$ is usually negligible for typical CMB experiments, so we have assumed $N_{\\ell}^{TE} = 0$. Then the Fisher matrix can be expressed as a sum over each $\\ell$, \n",
    "\n",
    "$$ F_{ij} = \\sum_{\\ell} \\frac{2 \\ell + 1}{2} f_{\\mathrm{sky}} \\mathrm{Tr}\\,\\left( \\mathbf{C}_{\\ell}^{-1} \\frac{\\partial \\mathbf{C}_{\\ell}}{\\partial \\theta_i} \\mathbf{C}_{\\ell}^{-1} \\frac{\\mathbf{C}_{\\ell}}{\\partial \\theta_j} \\right)$$\n",
    "\n",
    "Here, $f_{\\mathrm{sky}}$ refers to the fraction of the sky which the experiment covers, since you can think of the $a_{\\ell m}$ as having $(2 \\ell + 1)f_{\\mathrm{sky}}$ samples to work with. The ($\\mathrm{Tr}$) refers to the trace of the matrix. You might consult [Wu et al. 2014](https://arxiv.org/abs/1402.4108) if you need a second description."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Teach a cosmologist how to fish, and...\n",
    "\n",
    "Now you'll compute your own Fisher matrix. Let's start with a full-sky, cosmic variance-limited (CV-limited) experiment. In a CV-limited experiment, we have $N_{\\ell}^{TT} = 0$ and $N_{\\ell}^{EE} = 0$. This is the best-possible CMB experiment we could ever make.\n",
    "\n",
    "----\n",
    "## Assignment 2\n",
    "\n",
    "Using the `fiducial` CLASS object we computed earlier, print out the $2 \\times 2$ matrix $\\mathbf{C}_{\\ell}$ for $\\ell = 1000$. Convert it into a numpy array, and then print its inverse too."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.00000000e+00, 0.00000000e+00, 1.56306222e-10, ...,\n",
       "       4.73088092e-17, 4.72642305e-17, 4.72195451e-17])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# you might use for example...\n",
    "fiducial['tt']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Assignment 3\n",
    "Compute a single element of the Fisher matrix, where $\\theta_i = \\omega_b$ and $\\theta_j = h$. Use your derivative function you wrote earlier, and perform the matrix multiplications! Remember to sum over $\\ell$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Assignment 4\n",
    "Compute the full Fisher matrix over the set of parameters $\\{ \\omega_{cdm}, \\omega_b, h \\}$. That is, $F_{01}$ would refer to $\\theta_0 = \\omega_{cdm}$, and $\\theta_1 = \\omega_{b}$, and so on."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Assignment 5\n",
    "Print out your covariance! Recall that the best possible covariance of your parameters is the inverse of the Fisher matrix,\n",
    "\n",
    "$$ \\mathrm{Cov} = F^{-1} .$$\n",
    "\n",
    "This corresponds to the covariance matrix for a multivariate Gaussian which describes your parameters, also called the posterior. The ellipses one usually sees for Fisher forecasts are just 2D marginalized 1- and 2-$\\sigma$ contours. The 1-$\\sigma$ constraints marginalized over all other parameters (i.e. the error quoted in the $X \\pm \\sigma$) is the square root of the diagonal of the covariance matrix, i.e. $\\sigma_i = \\sqrt{\\mathrm{Cov}_{ii}}$. What's the error bar on your experiment for $h$?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Assignment 6\n",
    "Finally, we'll perform a more realistic forecast! We'll add in the noise terms $N_{\\ell}^{TT}$ and $N_{\\ell}^{EE}$. My favorite reference for this is [Wu et al. 2014](https://arxiv.org/abs/1402.4108). White noise curves can be generated with\n",
    "$$N_\\ell^{XX'} = s^2 \\text{exp} \\left( \\ell(\\ell+1) \\frac{\\theta^2_\\mathrm{FWHM}}{8 \\ln 2} \\right),$$\n",
    "where $s$ is the noise level in $\\mu K$-radians, and $\\theta_\\mathrm{FWHM}$ is the beam width in radians. A common forecasting prescription for Planck has temperature noise level $s_T = 33 \\mu \\mathrm{K}$-arcmin, polarization noise level $s_P = \\sqrt{2} s_T$, and the beam $\\theta_{FWHM} = 7$ arcmin. Importantly, you must remove a factor of $10^{-6} T_{cmb} / \\mathrm{Kelvin}$ to match the unitless output of CLASS, or vice versa. Make a forecast for Planck 2015!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "sT = 33 * (np.pi/60./180.)\n",
    "sP = sT * np.sqrt(2.)\n",
    "theta_FWHM = 7. * (np.pi/60./180.)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Once you've calculated a covariance matrix, it's straightforward to make triangle plots like below, as the math is just collapsing down the multivariate Gaussian you've computed. The implementation details for making pretty ellipses are in the very helpful [Coe 2009](https://arxiv.org/abs/0906.4123).\n",
    "\n",
    "<img src=\"../images/basic_output.png\" width=\"300\" height=\"300\" title=\"basic triangle plot\">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Bonus Material: Noise Plots\n",
    "\n",
    "It's sometimes useful to see visualize what's happening with Fisher matrices. Here I describe one way, where you plot the signal and noise curves. The binned variance (for a `binsize` perhaps of 100-200) can be written\n",
    "\n",
    "$$ \\sigma_{C_{\\ell}}^2 = \\frac{1}{N_{bin}}  \\frac{2}{(2 \\ell + 1)f_{sky}}  \\left(N_{\\ell}^2 + C_{\\ell}^2 \\right) , $$\n",
    "\n",
    "Basically the term in the $\\mathbf{C_{\\ell}}$ elements that go into $F_{ij}$. It is often convenient to plot the quantities $\\Delta C_{\\ell} / C_{\\ell}$ and $\\sigma_{C_{\\ell}} / C_{\\ell}$.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "-----\n",
    "## Some Dire Warnings\n",
    "\n",
    "Fisher forecasts are pretty crude estimates of the posterior. They will fail for all sorts of reasons, in particular\n",
    "\n",
    "1. **Noisy derivatives**. Fisher codes can unfortunately be sensitive to the step-size used in computing the derivative. In order to make robust forecasts, always make sure to check that your forecasts do not fluctuate too much in terms of the stepsize, and that your derivatives do not suffer from numerical error from using too small a step-size. This can sometimes be solved by increasing the precision parameters of the code which makes your theory curves.\n",
    "2. **Non-Gaussian posteriors and priors**. If your parameters relate to each other in banana shapes on a triangle plot, the Fisher forecast can be very wrong (probably at least a factor of 2). Fisher can in principle only make multivariate Gaussian posteriors. However, if you can reparametrize your model to a set of variables which do have a Gaussian posterior then you can make it work, see the appendix of [this](https://arxiv.org/abs/1806.10165). This can also be addressed by MCMC, which is more expensive computationally and comes with its own issues. For more about MCMC vs. Fisher, take a look at [Perotto et al. 2006](https://arxiv.org/abs/astro-ph/0606227).\n",
    "3. **Numerical issues with very degenerate parameters**. Suppose you have two parameters which are almost totally degenerate, like the dark energy equation of state parameter $w_0$, and the Hubble constant $H_0$ in the unlensed primary (thankfully lensing breaks this degeneracy for real data). This can result in extremely large numbers in your Fisher matrix, which you will be squaring due to the matrix multiplication. The danger here is in the matrix inversion, which can accumulate numerical errors due to the large number of operations involved.\n",
    "\n",
    "I will also advertise [my own Fisher code](https://github.com/xzackli/fishchips-public/) here, it's marginally better documented than other options and you can always contact me if something breaks!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}