VAR bugs: 1) lags=0, 2) select_order Raw

bugs__lags_and_selectorder.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas\n",
    "from statsmodels.tsa.vector_ar.var_model import VAR\n",
    "import data as dta  # statsmodels/datasets/macrodata/data.py\n",
    "from statsmodels.tsa.base.datetools import dates_from_str\n",
    "\n",
    "# prepare data\n",
    "mdata = dta.load_pandas().data\n",
    "iidata = dta.load_pandas();\n",
    "mdata = iidata.data\n",
    "dates = mdata[['year', 'quarter']].astype(int).astype(str)\n",
    "quarterly = dates[\"year\"] + \"Q\" + dates[\"quarter\"]\n",
    "from statsmodels.tsa.base.datetools import dates_from_str\n",
    "quarterly = dates_from_str(quarterly)\n",
    "mdata = mdata[[\"realcons\", \"realgdp\", \"realinv\"]]\n",
    "mdata.index = pandas.DatetimeIndex(quarterly)\n",
    "data = mdata"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Bug 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If `lags > 0`, everything is OK:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = VAR(data)\n",
    "est = model.fit(maxlags=4, trend=\"nc\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If `lags == 0` but there is some trend, everything is OK too:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = VAR(data)\n",
    "est = model.fit(maxlags=0, trend=\"c\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If `lags == 0` and there is no trend, we get an exception. The reason why we get this exception is that the matrix Z (see Lütkepohl, p. 70) has shape (203, 0) in this case, i.e. it doesn't have columns."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "math domain error",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-4-6ffc99743af7>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mVAR\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdata\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[0mest\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmaxlags\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrend\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m\"nc\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m/home/vegcev/D/TUG/Masterarbeit/GSoC/statsmodels-yogabonito/statsmodels/tsa/vector_ar/var_model.py\u001b[0m in \u001b[0;36mfit\u001b[1;34m(self, maxlags, method, ic, trend, verbose)\u001b[0m\n\u001b[0;32m    436\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnobs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mendog\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m-\u001b[0m \u001b[0mlags\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    437\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 438\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_estimate_var\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlags\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrend\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mtrend\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    439\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    440\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_estimate_var\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlags\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0moffset\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrend\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'c'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m/home/vegcev/D/TUG/Masterarbeit/GSoC/statsmodels-yogabonito/statsmodels/tsa/vector_ar/var_model.py\u001b[0m in \u001b[0;36m_estimate_var\u001b[1;34m(self, lags, offset, trend)\u001b[0m\n\u001b[0;32m    459\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    460\u001b[0m         \u001b[1;31m# Lutkepohl p75, about 5x faster than stated formula\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 461\u001b[1;33m         \u001b[0mparams\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlinalg\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlstsq\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mz\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0my_sample\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    462\u001b[0m         \u001b[0mresid\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0my_sample\u001b[0m \u001b[1;33m-\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdot\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mz\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mparams\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    463\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m/home/vegcev/.local/lib/python3.4/site-packages/numpy/linalg/linalg.py\u001b[0m in \u001b[0;36mlstsq\u001b[1;34m(a, b, rcond)\u001b[0m\n\u001b[0;32m   1887\u001b[0m     \u001b[0ma\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mbstar\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0m_to_native_byte_order\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0ma\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mbstar\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1888\u001b[0m     \u001b[0ms\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mzeros\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mm\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mreal_t\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1889\u001b[1;33m     \u001b[0mnlvl\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmax\u001b[0m\u001b[1;33m(\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mint\u001b[0m\u001b[1;33m(\u001b[0m \u001b[0mmath\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlog\u001b[0m\u001b[1;33m(\u001b[0m \u001b[0mfloat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mm\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;36m2.\u001b[0m \u001b[1;33m)\u001b[0m \u001b[1;33m)\u001b[0m \u001b[1;33m+\u001b[0m \u001b[1;36m1\u001b[0m \u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1890\u001b[0m     \u001b[0miwork\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mzeros\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m3\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mmin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mm\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mnlvl\u001b[0m\u001b[1;33m+\u001b[0m\u001b[1;36m11\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mmin\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mm\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfortran_int\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1891\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0misComplexType\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mt\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mValueError\u001b[0m: math domain error"
     ]
    }
   ],
   "source": [
    "model = VAR(data)\n",
    "est = model.fit(maxlags=0, trend=\"nc\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Bug 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`select_order()` doesn't take into account any deterministic terms (it always calls `_estimate_var()` without trend argument). So `_estimate_var()` uses the default trend `\"c\"`. If we add a `trend` parameter to `select_order()`, we can pass it to `_estimate_var()`. (In case of `trend == \"nc\"`, this will raise the same exception as above because `select_order()` will call `_estimate_var()` with `lags=0`.)"
   ]
  }
 ],
 "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": "3.4.3+"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}