generic_mle.ipynb

{
 "metadata": {
  "name": "generic_mle"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import numpy as np\n",
      "from scipy import stats\n",
      "import statsmodels.api as sm\n",
      "from statsmodels.base.model import GenericLikelihoodModel"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print sm.datasets.spector.NOTE"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "data = sm.datasets.spector.load_pandas()\n",
      "exog = sm.add_constant(data.exog, prepend=True)\n",
      "endog = data.endog"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "sm_probit = sm.Probit(endog, exog).fit()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "raw",
     "metadata": {},
     "source": [
      "* To create your own Likelihood Model, you just need to overwrite the loglike method."
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "class MyProbit(GenericLikelihoodModel):\n",
      "    def loglike(self, params):\n",
      "        exog = self.exog\n",
      "        endog = self.endog\n",
      "        q = 2 * endog - 1\n",
      "        return stats.norm.logcdf(q*np.dot(exog, params)).sum()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "my_probit = MyProbit(endog, exog).fit()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print sm_probit.params"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print sm_probit.cov_params()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print my_probit.params"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    },
    {
     "cell_type": "raw",
     "metadata": {},
     "source": [
      "You can get the variance-covariance of the parameters. Notice that we didn't have to provide Hessian or Score functions."
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print my_probit.cov_params()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}