DAGs.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3 Directed Acyclic Graphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import statsmodels.api as sm\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.1.5 Discrimination and collider bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# compare to collider_discrimination\n",
    "\n",
    "population_size = 10000\n",
    "\n",
    "# create a dataframe for the observations\n",
    "df = pd.DataFrame(index = range(population_size))\n",
    "\n",
    "# Indicator variable for female\n",
    "df['female'] = np.random.binomial(1, .5, population_size)\n",
    "\n",
    "# Innate ability is independent of gender\n",
    "df['ability'] = np.random.normal(size = population_size)\n",
    "\n",
    "# All women experience discrimination \n",
    "df['discrimination'] = df['female']\n",
    "\n",
    "# Data generating process\n",
    "df['occupation'] = 1 + 2*df.ability + 0*df.female - 2*df.discrimination + np.random.normal(size = population_size)\n",
    "df['wage'] = 1 - 1*df.discrimination + df.occupation + 2*df.ability + np.random.normal(size = population_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Biased Unconditional"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                            OLS Regression Results                            \n",
      "==============================================================================\n",
      "Dep. Variable:                   wage   R-squared:                       0.116\n",
      "Model:                            OLS   Adj. R-squared:                  0.116\n",
      "Method:                 Least Squares   F-statistic:                     1313.\n",
      "Date:                Sat, 13 Mar 2021   Prob (F-statistic):          2.84e-270\n",
      "Time:                        10:54:37   Log-Likelihood:                -28594.\n",
      "No. Observations:               10000   AIC:                         5.719e+04\n",
      "Df Residuals:                    9998   BIC:                         5.721e+04\n",
      "Df Model:                           1                                         \n",
      "Covariance Type:            nonrobust                                         \n",
      "==============================================================================\n",
      "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
      "------------------------------------------------------------------------------\n",
      "const          2.0447      0.060     34.002      0.000       1.927       2.163\n",
      "female        -3.0611      0.084    -36.237      0.000      -3.227      -2.895\n",
      "==============================================================================\n",
      "Omnibus:                        3.403   Durbin-Watson:                   1.975\n",
      "Prob(Omnibus):                  0.182   Jarque-Bera (JB):                3.396\n",
      "Skew:                           0.034   Prob(JB):                        0.183\n",
      "Kurtosis:                       2.940   Cond. No.                         2.63\n",
      "==============================================================================\n",
      "\n",
      "Warnings:\n",
      "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n"
     ]
    }
   ],
   "source": [
    "X = df[['female']]\n",
    "y = df.wage\n",
    "\n",
    "X = sm.add_constant(X) # adding a constant\n",
    "\n",
    "model = sm.OLS(y, X).fit()\n",
    "\n",
    "print_model = model.summary()\n",
    "print(print_model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Biased Conditional\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                            OLS Regression Results                            \n",
      "==============================================================================\n",
      "Dep. Variable:                   wage   R-squared:                       0.909\n",
      "Model:                            OLS   Adj. R-squared:                  0.909\n",
      "Method:                 Least Squares   F-statistic:                 4.983e+04\n",
      "Date:                Sat, 13 Mar 2021   Prob (F-statistic):               0.00\n",
      "Time:                        10:51:19   Log-Likelihood:                -17236.\n",
      "No. Observations:               10000   AIC:                         3.448e+04\n",
      "Df Residuals:                    9997   BIC:                         3.450e+04\n",
      "Df Model:                           2                                         \n",
      "Covariance Type:            nonrobust                                         \n",
      "==============================================================================\n",
      "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
      "------------------------------------------------------------------------------\n",
      "const          0.2157      0.020     10.633      0.000       0.176       0.255\n",
      "female         0.5666      0.030     19.021      0.000       0.508       0.625\n",
      "occupation     1.7929      0.006    294.840      0.000       1.781       1.805\n",
      "==============================================================================\n",
      "Omnibus:                        1.451   Durbin-Watson:                   1.977\n",
      "Prob(Omnibus):                  0.484   Jarque-Bera (JB):                1.450\n",
      "Skew:                           0.005   Prob(JB):                        0.484\n",
      "Kurtosis:                       2.942   Cond. No.                         6.19\n",
      "==============================================================================\n",
      "\n",
      "Warnings:\n",
      "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n"
     ]
    }
   ],
   "source": [
    "X = df[['female', 'occupation']]\n",
    "y = df.wage\n",
    "\n",
    "X = sm.add_constant(X) # adding a constant\n",
    "\n",
    "model = sm.OLS(y, X).fit()\n",
    "\n",
    "print_model = model.summary()\n",
    "print(print_model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Unbiased Conditional"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                            OLS Regression Results                            \n",
      "==============================================================================\n",
      "Dep. Variable:                   wage   R-squared:                       0.950\n",
      "Model:                            OLS   Adj. R-squared:                  0.950\n",
      "Method:                 Least Squares   F-statistic:                 6.283e+04\n",
      "Date:                Sat, 13 Mar 2021   Prob (F-statistic):               0.00\n",
      "Time:                        10:51:47   Log-Likelihood:                -14269.\n",
      "No. Observations:               10000   AIC:                         2.855e+04\n",
      "Df Residuals:                    9996   BIC:                         2.857e+04\n",
      "Df Model:                           3                                         \n",
      "Covariance Type:            nonrobust                                         \n",
      "==============================================================================\n",
      "                 coef    std err          t      P>|t|      [0.025      0.975]\n",
      "------------------------------------------------------------------------------\n",
      "const          1.0118      0.017     57.875      0.000       0.977       1.046\n",
      "female        -1.0118      0.028    -35.819      0.000      -1.067      -0.956\n",
      "occupation     1.0011      0.010    101.203      0.000       0.982       1.020\n",
      "ability        1.9951      0.022     89.989      0.000       1.952       2.039\n",
      "==============================================================================\n",
      "Omnibus:                        2.421   Durbin-Watson:                   1.982\n",
      "Prob(Omnibus):                  0.298   Jarque-Bera (JB):                2.451\n",
      "Skew:                          -0.034   Prob(JB):                        0.294\n",
      "Kurtosis:                       2.963   Cond. No.                         9.40\n",
      "==============================================================================\n",
      "\n",
      "Warnings:\n",
      "[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n"
     ]
    }
   ],
   "source": [
    "X = df[['female', 'occupation', 'ability']]\n",
    "y = df.wage\n",
    "\n",
    "X = sm.add_constant(X) # adding a constant\n",
    "\n",
    "model = sm.OLS(y, X).fit()\n",
    "\n",
    "print_model = model.summary()\n",
    "print(print_model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3.1.6 Sample selection and collider bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# compare to moviestar\n",
    "\n",
    "obs = 2500\n",
    "df = pd.DataFrame(index = range(obs))\n",
    "\n",
    "df['beauty'] = np.random.normal(size = obs)\n",
    "df['talent'] = np.random.normal(size = obs)\n",
    "\n",
    "# create the collider variable\n",
    "df['score'] = df.beauty + df.talent\n",
    "df['star'] = df.score >= df.score.quantile(.85)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Make scatter plots\n",
    "\n",
    "stars = df[df.star == True]\n",
    "nonstars = df[df.star == False]\n",
    "\n",
    "fig = plt.figure()\n",
    "\n",
    "# Non-stars\n",
    "ax1 = fig.add_subplot(2,2,1)\n",
    "ax1.scatter(nonstars.talent, nonstars.beauty, s = 2)\n",
    "ax1.set_title(\"Non-stars\")\n",
    "\n",
    "# Stars\n",
    "ax2 = fig.add_subplot(2,2,2)\n",
    "ax2.scatter(stars.talent, stars.beauty, s = 2)\n",
    "ax2.set_title('Stars')\n",
    "\n",
    "ax3 = fig.add_subplot(2,2,3)\n",
    "ax3.scatter(df.talent, df.beauty, s = 2)\n",
    "ax3.set_title(\"All actors\")\n",
    "\n",
    "# Set axes limits\n",
    "for ax_ in ax0, ax1, ax2:\n",
    "    ax_.set_xlim([-4,4])\n",
    "    ax_.set_ylim([-4,4])\n",
    "\n",
    "# Label axes for entire grid\n",
    "fig.text(0.5, -.01, 'Talent', ha='center', va='center')\n",
    "fig.text(-0.01, 0.51, 'Beauty', ha='center', va='center', rotation='vertical')\n",
    "    \n",
    "    \n",
    "fig.tight_layout()"
   ]
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}