Updated AIPTW

awoziji · Mar 23, 2019 · 09cfd77 · 09cfd77
1 parent b20109f
commit 09cfd77
Showing 1 changed file with 21 additions and 18 deletions.
diff --git a/3_Epidemiology_Analysis/c_causal_inference/1_time-fixed-treatments/5_AIPTW.ipynb b/3_Epidemiology_Analysis/c_causal_inference/1_time-fixed-treatments/5_AIPTW.ipynb
@@ -78,7 +78,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -95,7 +95,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [
     {
@@ -113,8 +113,8 @@
       "Model Family:                Binomial   Df Model:                            8\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -206.06\n",
-      "Date:                Mon, 11 Mar 2019   Deviance:                       412.12\n",
-      "Time:                        08:24:05   Pearson chi2:                     510.\n",
+      "Date:                Sat, 23 Mar 2019   Deviance:                       412.12\n",
+      "Time:                        18:18:31   Pearson chi2:                     510.\n",
       "No. Iterations:                     5   Covariance Type:             nonrobust\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
@@ -148,7 +148,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 4,
    "metadata": {},
    "outputs": [
     {
@@ -166,8 +166,8 @@
       "Model Family:                Binomial   Df Model:                            9\n",
       "Link Function:                  logit   Scale:                          1.0000\n",
       "Method:                          IRLS   Log-Likelihood:                -202.85\n",
-      "Date:                Mon, 11 Mar 2019   Deviance:                       405.71\n",
-      "Time:                        08:25:18   Pearson chi2:                     535.\n",
+      "Date:                Sat, 23 Mar 2019   Deviance:                       405.71\n",
+      "Time:                        18:18:31   Pearson chi2:                     535.\n",
       "No. Iterations:                     6   Covariance Type:             nonrobust\n",
       "==============================================================================\n",
       "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
@@ -202,19 +202,22 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 5,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "RD: -0.06857139263314216\n",
-      "RR: 0.5844630051369846\n",
+      "RD: -0.08485106054489719\n",
+      "RR: 0.5319812234515429\n",
+      "======================================================================\n",
+      "           Augment Inverse Probability of Treatment Weights           \n",
+      "======================================================================\n",
+      "Risk Difference:    -0.0849\n",
       "----------------------------------------------------------------------\n",
-      "Risk Difference:  -0.0686\n",
-      "Risk Ratio:  0.5845\n",
-      "----------------------------------------------------------------------\n"
+      "Risk Ratio:         0.532\n",
+      "======================================================================\n"
      ]
     }
    ],
@@ -231,23 +234,23 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Interpreting the risk difference, we would conclude that had everyone in our cohort been treated with ART, the risk of all-cause mortality would have been 6.9% points lower than had no one been treated.\n",
+    "Interpreting the risk difference, we would conclude that had everyone in our cohort been treated with ART, the risk of all-cause mortality would have been 8.5% points lower than had no one been treated.\n",
     "\n",
     "### Confidence Intervals\n",
     "To obtain correct confidence intervals, we need to use a bootstrap procedure. Influence curves are an alternative, but not currently available"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": 6,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "95% LCL: -0.12324970574432371\n",
-      "95% UCL -0.005298084807482341\n"
+      "95% LCL: -0.14683185675036045\n",
+      "95% UCL -0.014802615734624436\n"
      ]
     }
    ],
@@ -272,7 +275,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Under the counterfactual of everyone receiving treatment with ART, the risk of all-cause mortality was 6.9% points lower (95% CL: ) than the counterfactual where no one had been treated.\n",
+    "Under the counterfactual of everyone receiving treatment with ART, the risk of all-cause mortality was 8.5% points lower (95% CL: -0.15, -0.01) than the counterfactual where no one had been treated.\n",
     "\n",
     "# Conclusion\n",
     "In this tutorial, I introduced the concept of doubly-robust estimators and detailed augmented-IPTW. I demonstrated estimation with `AIPTW` using *zEpid* and how to obtain confidence intervals. Please view other tutorials for information on other functionality within *zEpid*\n",