[FINALIZED] nb6 is ready for prime time

Author: ericmjl

notebooks/07-hierarchical-baseball.ipynb …6-instructor-hierarchical-baseball.ipynbnotebooks/07-hierarchical-baseball.ipynb renamed to notebooks/06-instructor-hierarchical-baseball.ipynb notebooks/07-hierarchical-baseball.ipynb renamed to notebooks/06-instructor-hierarchical-baseball.ipynb

@@ -39,9 +39,7 @@
     "1. What metric would you rank by? \n",
     "1. Would your metric be reasonable for rookie players?\n",
     "\n",
-    "### Exercise\n",
-    "\n",
-    "Read in the data, which is located at `../data/baseballdb/core/Batting.csv`."
+    "Let's read in the data."
    ]
   },
   {
@@ -129,6 +127,13 @@
     "    like = pm.Binomial('likelihood', n=df['AB'], p=thetas, observed=df['H'])"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Now, sample from the posterior"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -139,6 +144,13 @@
     "    baseline_trace = pm.sample(2000)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Plot the traceplots to check for convergence."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -149,14 +161,21 @@
     "despine_traceplot(traceplot)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Visualize the posterior distribution using forest plots."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
     "ylabels = ylabels = \"AB: \" + df['AB'].astype(str) + ', H: ' + df['H'].astype('str')\n",
-    "pm.forestplot(baseline_trace, ylabels=ylabels)"
+    "pm.forestplot(baseline_trace, ylabels=ylabels, xlim=[0, 1])"
    ]
   },
   {
@@ -219,6 +238,13 @@
     "    like = pm.Binomial('like', n=df['AB'], p=thetas, observed=df['H'])"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Sample from the posterior."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -229,6 +255,13 @@
     "    trace = pm.sample(2000, nuts_kwargs={'target_accept': 0.95})"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Visualize the trace plots to check for convergence."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -239,14 +272,21 @@
     "despine_traceplot(ax_arr)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Visualize the posterior distributions using forestplots."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
     "ylabels = \"AB: \" + df['AB'].astype(str) + ', H: ' + df['H'].astype('str')\n",
-    "pm.forestplot(trace, varnames=['thetas'], ylabels=ylabels)"
+    "pm.forestplot(trace, varnames=['thetas'], ylabels=ylabels, xlim=[0, 1])"
    ]
   },
   {
@@ -301,6 +341,16 @@
     "- Under the assumption of hierarchical structure holding true, we will not get estimates that one may consider to be absurd (e.g. long-run batting probability estimated to be zero or one, on the basis onf few observations)."
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Further Exercises\n",
+    "\n",
+    "- Change the priors. How do they change the posteriors?\n",
+    "- Change the parameterization of the model. How do they change your inference?"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},