case study v1

Author: AIgoGracia

book/cate_and_policy/policy_learning.ipynb

@@ -500,11 +500,119 @@
     "print(\"ANALYSIS COMPLETE\")\n",
     "print(\"=\" * 70)"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "import warnings\n",
+    "warnings.filterwarnings('ignore')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Loading data...\n",
+      "Data loaded: 29726 observations\n",
+      "Treatment variable: w\n",
+      "Outcome variable: y\n",
+      "Covariates: age, polviews, income, educ, marital, sex\n",
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(\"Loading data...\")\n",
+    "# Read in data\n",
+    "url = \"https://docs.google.com/uc?id=1AQva5-vDlgBcM_Tv9yrO8yMYRfQJgqo_&export=download\"\n",
+    "data = pd.read_csv(url)\n",
+    "n = len(data)\n",
+    "\n",
+    "# NOTE: We'll invert treatment and control, compared to previous chapters\n",
+    "data['w'] = 1 - data['w']\n",
+    "\n",
+    "# Treatment is the wording of the question:\n",
+    "# 'does the gov't spend too much on 'assistance to the poor' (control: 0)\n",
+    "# 'does the gov't spend too much on \"welfare\"?' (treatment: 1)\n",
+    "treatment = 'w'\n",
+    "\n",
+    "# Outcome: 1 for 'yes', 0 for 'no'\n",
+    "outcome = 'y'\n",
+    "\n",
+    "# Additional covariates\n",
+    "covariates = ['age', 'polviews', 'income', 'educ', 'marital', 'sex']\n",
+    "\n",
+    "print(f\"Data loaded: {n} observations\")\n",
+    "print(f\"Treatment variable: {treatment}\")\n",
+    "print(f\"Outcome variable: {outcome}\")\n",
+    "print(f\"Covariates: {', '.join(covariates)}\")\n",
+    "print()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "======================================================================\n",
+      "METHOD 1: Simple Mean-Based Estimation\n",
+      "(Only valid in randomized setting)\n",
+      "======================================================================\n",
+      "\n",
+      "Policy: Treat if polviews <= 4 OR age > 50\n",
+      "Proportion treated under policy: 0.773\n",
+      "Value estimate: 0.3457179812 Std. Error: 0.0038947280\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(\"=\" * 70)\n",
+    "print(\"METHOD 1: Simple Mean-Based Estimation\")\n",
+    "print(\"(Only valid in randomized setting)\")\n",
+    "print(\"=\" * 70)\n",
+    "\n",
+    "# Extract variables\n",
+    "X = data[covariates].values\n",
+    "Y = data[outcome].values\n",
+    "W = data[treatment].values\n",
+    "\n",
+    "# Define policy: treat if polviews <= 4 (liberal/moderate) OR age > 50\n",
+    "pi = (data['polviews'].values <= 4) | (data['age'].values > 50)\n",
+    "A = pi == 1\n",
+    "\n",
+    "# Calculate value estimate\n",
+    "value_estimate = np.mean(Y[A & (W==1)]) * np.mean(A) + \\\n",
+    "                 np.mean(Y[~A & (W==0)]) * np.mean(~A)\n",
+    "\n",
+    "# Calculate standard error\n",
+    "value_stderr = np.sqrt(\n",
+    "    np.var(Y[A & (W==1)]) / np.sum(A & (W==1)) * np.mean(A)**2 + \n",
+    "    np.var(Y[~A & (W==0)]) / np.sum(~A & (W==0)) * np.mean(~A)**2\n",
+    ")\n",
+    "\n",
+    "print(f\"\\nPolicy: Treat if polviews <= 4 OR age > 50\")\n",
+    "print(f\"Proportion treated under policy: {np.mean(A):.3f}\")\n",
+    "print(f\"Value estimate: {value_estimate:.10f} Std. Error: {value_stderr:.10f}\")"
+   ]
   }
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "base",
+   "display_name": "Python 3",
    "language": "python",
    "name": "python3"
   },
@@ -518,7 +626,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.18"
+   "version": "3.11.9"
   }
  },
  "nbformat": 4,