Comparison of performance using statsmodels.formula.api and statsmodels.api.

statsmodels_comparison.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:cda612edbb49dc07450f70b9797e0bf67b9d6ae2e7bcb0cf46aa2645bdc961e1"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import pandas as pd\n",
      "import numpy as np\n",
      "import statsmodels.api as sm\n",
      "import statsmodels.formula.api as smf"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "patsy_exp = 'a ~ np.exp(b) + np.log(c)'"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 41
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "num = 5000000\n",
      "df = pd.DataFrame({'a': np.linspace(10, num, num), \n",
      "                   'b': np.log(np.linspace(500, num, num)),\n",
      "                   'c': np.exp(np.linspace(1, 100, num))})"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 60
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%timeit fresult = smf.ols(formula=patsy_exp, data=df).fit()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "1 loops, best of 3: 2.21 s per loop\n"
       ]
      }
     ],
     "prompt_number": 61
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%%timeit\n",
      "df2 = pd.DataFrame({'a': df.a,\n",
      "                    'b': df.b.apply(np.exp),\n",
      "                    'c': df.c.apply(np.log)})\n",
      "result = sm.OLS(df2.a, df2[['b', 'c']]).fit()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "1 loops, best of 3: 664 ms per loop\n"
       ]
      }
     ],
     "prompt_number": 62
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "test_df = pd.DataFrame({'b': [2], 'c': [3]})\n",
      "test_df2 = pd.DataFrame({'b': np.exp([2]), 'c': np.log([3])})"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 57
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "smf.ols(formula=patsy_exp, data=df).fit().predict(test_df)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 58,
       "text": [
        "array([-530.96187943])"
       ]
      }
     ],
     "prompt_number": 58
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "sm.OLS(df2.a, df2[['b', 'c']]).fit().predict(test_df2)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 59,
       "text": [
        "array([-589.93971495])"
       ]
      }
     ],
     "prompt_number": 59
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}