PyMC out of sample predictions.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "149de8b6",
   "metadata": {},
   "source": [
    "# Introduction\n",
    "\n",
    "A cool thing about hierarchical models is that its easy to predict out of sample - i.e. if you want to make a prediction on a new zipcode, just sample from the state's distribution (composed of the state average and variance across zip codes in that state)\n",
    "\n",
    "In pymc3, it's somewhat easy to accomplish this, but not as straightforward as we'd hope. This blog post will show a trick that lets you easily predict out of sample, and will reduce some of the overhead that comes from writing alot of custom prediction functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "06164097",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt \n",
    "import seaborn as sns\n",
    "import pymc3 as pm\n",
    "import arviz as az\n",
    "SEED = 99"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7edbb346",
   "metadata": {},
   "source": [
    "# Simulating data\n",
    "We'll simulate a 2 level hierarchical model - for interpretability, we'll set it up as a state > zipcode model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "072281cc",
   "metadata": {},
   "outputs": [],
   "source": [
    "def gen_zipcode(size=1):\n",
    "    '''Ugly way to generate fake zip codes'''\n",
    "    codes = []\n",
    "    for _ in range(int(2*size) + 10 ):\n",
    "        code = ''\n",
    "        for _ in range(5):\n",
    "            code += np.random.choice(range(9)).astype(str)\n",
    "        codes.append(code)\n",
    "        \n",
    "    return np.array(list(set(codes))[:size])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "06c0a151",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Simulate data\n",
    "np.random.seed(SEED)\n",
    "N_states = 10\n",
    "N_zips_per_state = 15\n",
    "\n",
    "# Coefficients\n",
    "mu_country_true = 0\n",
    "sig_state_true = 5\n",
    "mu_state_true = pm.Normal.dist(mu_country_true, sig_state_true)\n",
    "states_true = mu_state_true.random(size=N_states)\n",
    "\n",
    "sig_zip_true = 3\n",
    "mu_zipcode_true = pm.Normal.dist(states_true, sig_zip_true)\n",
    "zipcode_true = mu_zipcode_true.random(size=N_zips_per_state)\n",
    "\n",
    "# Indexes for observed dataa\n",
    "state_idx = np.random.choice(np.arange(N_states), size=1000)\n",
    "zip_idx = np.random.choice(np.arange(N_zips_per_state), size=1000)\n",
    "\n",
    "# observed data\n",
    "sig_obs_true = 2\n",
    "y_true = np.random.normal( zipcode_true[zip_idx, state_idx], sig_obs_true)\n",
    "\n",
    "# Labels\n",
    "states = np.sort(np.array([\"CA\", \"TX\", \"FL\", \"GA\", \"AL\", \"ME\", \"MA\", \"CT\", \"NY\", \"NJ\"]))\n",
    "zip_codes = np.sort(gen_zipcode(N_states*N_zips_per_state))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "cfd6f4c5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Change zipcode index to 0-1000 instead of 0-100 by \n",
    "# making these fake zipcodes nested within states\n",
    "zip_idx_adj = (pd.Series( state_idx*N_zips_per_state + zip_idx ).astype(\"str\") \n",
    "              ).astype(int).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "096b1fe2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>y</th>\n",
       "      <th>state</th>\n",
       "      <th>zipcode</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6.953332</td>\n",
       "      <td>ME</td>\n",
       "      <td>61756</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8.789451</td>\n",
       "      <td>FL</td>\n",
       "      <td>28766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6.810024</td>\n",
       "      <td>FL</td>\n",
       "      <td>26505</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7.922506</td>\n",
       "      <td>ME</td>\n",
       "      <td>63467</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>13.250428</td>\n",
       "      <td>CA</td>\n",
       "      <td>11450</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>-6.689709</td>\n",
       "      <td>AL</td>\n",
       "      <td>04463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>-7.306024</td>\n",
       "      <td>TX</td>\n",
       "      <td>85822</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>-0.031004</td>\n",
       "      <td>CT</td>\n",
       "      <td>23215</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.307124</td>\n",
       "      <td>GA</td>\n",
       "      <td>40461</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.175902</td>\n",
       "      <td>GA</td>\n",
       "      <td>35720</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           y state zipcode\n",
       "0   6.953332    ME   61756\n",
       "1   8.789451    FL   28766\n",
       "2   6.810024    FL   26505\n",
       "3   7.922506    ME   63467\n",
       "4  13.250428    CA   11450\n",
       "5  -6.689709    AL   04463\n",
       "6  -7.306024    TX   85822\n",
       "7  -0.031004    CT   23215\n",
       "8   0.307124    GA   40461\n",
       "9   0.175902    GA   35720"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# organizing the data\n",
    "df = (\n",
    "    pd.DataFrame({\n",
    "        \"y\":y_true,\n",
    "        \"state\":states[state_idx],\n",
    "        \"zipcode\":zip_codes[zip_idx_adj]\n",
    "    })\n",
    ")\n",
    "\n",
    "df.head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4f08df7c",
   "metadata": {},
   "source": [
    "## Using categorical variables for convenience\n",
    "\n",
    "Categorical variables are a somewhat new feature of pandas - they can store categories that aren't in the observed data, and are an easy replacement for `pd.factorize()`\n",
    "\n",
    "We can use these to trick pymc into thinking there's a category with no observed data, and pymc ends up assigning the global distribution to that unobserved category, which we can simply reference in the future for any time we want to make a prediction on out of sample data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "d9011637",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>y</th>\n",
       "      <th>state</th>\n",
       "      <th>zipcode</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6.953332</td>\n",
       "      <td>ME</td>\n",
       "      <td>61756</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>8.789451</td>\n",
       "      <td>FL</td>\n",
       "      <td>28766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6.810024</td>\n",
       "      <td>FL</td>\n",
       "      <td>26505</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>7.922506</td>\n",
       "      <td>ME</td>\n",
       "      <td>63467</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>13.250428</td>\n",
       "      <td>CA</td>\n",
       "      <td>11450</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           y state zipcode\n",
       "0   6.953332    ME   61756\n",
       "1   8.789451    FL   28766\n",
       "2   6.810024    FL   26505\n",
       "3   7.922506    ME   63467\n",
       "4  13.250428    CA   11450"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Convert to categorical and add a `global` category\n",
    "df = df.assign(state = pd.Categorical(df.state).add_categories(\"out_of_sample\"))\\\n",
    "    .assign(zipcode = pd.Categorical(df.zipcode).add_categories(\"out_of_sample\"))\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d1e72fe3",
   "metadata": {},
   "source": [
    "# Fitting the model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "602dbd16",
   "metadata": {},
   "source": [
    "We'll use the `codes` from the categorical columns to index our model coefficients, and we'll use the `categories` as coordinates for the model to map names to"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "f5810711",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['AL', 'CA', 'CT', 'FL', 'GA', 'MA', 'ME', 'NJ', 'NY', 'TX',\n",
       "       'out_of_sample'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coords={\n",
    "    \"state\":df.state.cat.categories,\n",
    "    \"zipcode\":df.zipcode.cat.categories\n",
    "}\n",
    "\n",
    "# Peak at state labels\n",
    "coords[\"state\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f1069718",
   "metadata": {},
   "outputs": [],
   "source": [
    "# # This model could have sampled more efficiently, commenting it out\n",
    "\n",
    "# with pm.Model(coords=coords) as model:\n",
    "    \n",
    "#     # Hyperpriors\n",
    "#     mu_country = pm.Normal(\"mu_country\", 0, 3)\n",
    "#     sig_country = pm.Exponential(\"sig_country\", 2.5)\n",
    "    \n",
    "#     # priors\n",
    "#     sig_zipcode = pm.Exponential(\"sig_zipcode\", 1)\n",
    "#     sig = pm.Exponential(\"sig\", 1)\n",
    "    \n",
    "#     # Hierarchical coefficients\n",
    "#     mu_state = pm.Normal(\"mu_state\", mu_country, sig_country, dims=\"state\")\n",
    "#     mu_zipcode = pm.Normal(\"mu_zipcode\", mu_state, sig_zipcode, dims=(\"zipcode\", \"state\"))\n",
    "    \n",
    "#     # Observational model\n",
    "#     y = pm.Normal(\"y\", mu_zipcode[zip_, state_], sig, observed=df.y)\n",
    "    \n",
    "#     # Fit\n",
    "#     trace = pm.sample(target_accept=0.9, return_inferencedata=True, random_seed=SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c6095e44",
   "metadata": {},
   "source": [
    "The model above had some difficulties sampling, so we'll try a noncentered parameterization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "6d4527b0",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Auto-assigning NUTS sampler...\n",
      "Initializing NUTS using jitter+adapt_diag...\n",
      "Multiprocess sampling (4 chains in 4 jobs)\n",
      "NUTS: [σ_mu_zipcode, Δ_mu_zipcode, σ_mu_state, Δ_mu_state, sig, mu_country]\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='8000' class='' max='8000' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      100.00% [8000/8000 00:48<00:00 Sampling 4 chains, 0 divergences]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 60 seconds.\n",
      "The number of effective samples is smaller than 25% for some parameters.\n"
     ]
    }
   ],
   "source": [
    "def hierarchical_normal(name, μ, dims):\n",
    "    '''Adapted from Austin Rochford'''\n",
    "    Δ = pm.Normal('Δ_{}'.format(name), 0., 1., dims=dims)\n",
    "    σ = pm.Exponential('σ_{}'.format(name), 2.5)\n",
    "    return pm.Deterministic(name, μ + Δ * σ, dims=dims)\n",
    "\n",
    "\n",
    "with pm.Model(coords=coords) as model_nc:\n",
    "    \n",
    "    # Observed Data tracking\n",
    "    state_ = pm.Data(\"state_\", df.state.cat.codes)\n",
    "    zip_ = pm.Data(\"zip_\", df.zipcode.cat.codes)\n",
    "    obs = pm.Data(\"obs\", df.y)\n",
    "\n",
    "    # Hyperprior\n",
    "    mu_country = pm.Normal(\"mu_country\", 0, 1)\n",
    "    \n",
    "    # Prior\n",
    "    sig = pm.Exponential(\"sig\", 1)\n",
    "    \n",
    "    # Hierarchical coefficients\n",
    "    mu_state = hierarchical_normal(\"mu_state\", μ=mu_country, dims=\"state\")\n",
    "    mu_zipcode = hierarchical_normal(\"mu_zipcode\", μ=mu_state, dims=(\"zipcode\", \"state\") )\n",
    "    \n",
    "    # Observational model\n",
    "    y = pm.Normal(\"y\", mu_zipcode[zip_, state_], sig, observed=obs)\n",
    "    \n",
    "    # Fit \n",
    "    trace_nc = pm.sample(target_accept=0.9, return_inferencedata=True, random_seed=SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe5cd328",
   "metadata": {},
   "source": [
    "Let's take a peak at at the state level coefficients"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "f4f1174b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([<AxesSubplot:title={'center':'94.0% HDI'}>], dtype=object)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x604.8 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "pm.plot_forest(trace_nc, var_names=[\"mu_state\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8c22f6e0",
   "metadata": {},
   "source": [
    "Great, that `out of sample` variable seems to represent the global distribution across states - i.e. if we were to make a prediction for a new state we'd use that distribtion. We should still confirm it to be safe\n",
    "\n",
    "We'll check for zip codes below as well, looking at Maine specifically"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "75cb5458",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x748.8 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "ME_zips = df.loc[lambda d: d.state==\"ME\"].zipcode.tolist() + [\"out_of_sample\"]\n",
    "pm.plot_forest(trace_nc, var_names=[\"mu_zipcode\"],\n",
    "              coords={\"state\":[\"ME\"], \"zipcode\":np.unique( ME_zips )});"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ebd88299",
   "metadata": {},
   "source": [
    "As we can see, the `out_of_sample` variable has a sampled value despite there being no observed data for it\n",
    "\n",
    "Now the question is, does this align with how we'd predict new data?\n",
    "\n",
    "Let's try calculating coefficients out of sample by hand and see if it aligns with the `out_of_sample` values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "ca979ea0",
   "metadata": {},
   "outputs": [],
   "source": [
    "post = trace_nc.posterior\n",
    "\n",
    "# Calculate out of sample state means by drawing from global distribution\n",
    "mu_country = post[\"mu_country\"].values.reshape(4000,-1)\n",
    "σ_state = post[\"σ_mu_state\"].values.reshape(4000,-1)\n",
    "mu_state = np.random.normal(mu_country, σ_state)\n",
    "\n",
    "# calculate out of sample mu by drawing from out of sample state prediction above\n",
    "σ_zipcode = post[\"σ_mu_zipcode\"].values.reshape(4000,-1)\n",
    "mu_zipcode = np.random.normal(mu_state, σ_zipcode)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "47ed22e0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot comparison\n",
    "fig, ax = plt.subplots(2,1,figsize=(10,7),sharex=True)\n",
    "\n",
    "# Plot 1\n",
    "ax[0].set_title(\"out of sample state estimate compared to indexing trick\")\n",
    "\n",
    "# Ground truth\n",
    "sns.histplot( mu_state_true.random(size=4000), \n",
    "             alpha=0.5, \n",
    "             color=\"C2\",\n",
    "             label=\"True Value\",\n",
    "             stat=\"probability\",\n",
    "            ax=ax[0])\n",
    "\n",
    "# Manual calculation from fitted model\n",
    "sns.histplot( mu_state.ravel(), \n",
    "             alpha=0.5, \n",
    "             label=\"Manual Prediction\",\n",
    "             stat=\"probability\",\n",
    "            ax=ax[0]);\n",
    "\n",
    "# Indexing trick\n",
    "sns.histplot( post[\"mu_state\"].sel({\"state\":[\"out_of_sample\"]}).values.ravel(), \n",
    "             alpha=0.5, \n",
    "             color=\"C1\", \n",
    "             label=\"Indexing Trick Prediction\",\n",
    "             stat=\"probability\",\n",
    "            ax=ax[0])\n",
    "ax[0].legend()\n",
    "\n",
    "\n",
    "# Plot 2\n",
    "ax[1].set_title(\"out of sample zipcode estimate compared to indexing trick\")\n",
    "\n",
    "# Ground truth\n",
    "sns.histplot( pm.Normal.dist(mu_state_true.random(size=4000), sig_zip_true).random(size=4000), \n",
    "             alpha=0.5, \n",
    "             color=\"C2\",\n",
    "             label=\"True Value\",\n",
    "             stat=\"probability\",\n",
    "            ax=ax[1])\n",
    "\n",
    "# Manual calculation from fitted model\n",
    "sns.histplot( mu_zipcode.ravel(), \n",
    "             alpha=0.5, \n",
    "             label=\"Manual Prediction\",\n",
    "             stat=\"probability\",\n",
    "            ax=ax[1])\n",
    "\n",
    "# Indexing trick\n",
    "sns.histplot( post[\"mu_zipcode\"].sel({\"state\":[\"out_of_sample\"], \"zipcode\":[\"out_of_sample\"]}).values.ravel(), \n",
    "             alpha=0.5, \n",
    "             color=\"C1\", \n",
    "             label=\"Indexing Trick Prediction\",\n",
    "             stat=\"probability\",\n",
    "            ax=ax[1])\n",
    "ax[1].legend()\n",
    "ax[0].set_xlabel(\"coef value\")\n",
    "ax[1].set_xlabel(\"coef value\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7649fc25",
   "metadata": {},
   "source": [
    "As we can see above, the indexing trick recovers the true parameters - this can make prediction MUCH easier, especially if we change model paramaterization"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "713a4d2e",
   "metadata": {},
   "source": [
    "# Predicting out of sample\n",
    "\n",
    "Let's go ahead and actually make prediction now - we'll make predictions for the following data below\n",
    " * The first example is in sample\n",
    " * The second example is in sample for state, out of sample for zipcode\n",
    " * The third example is out of sample entirely"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "cb533529",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>state</th>\n",
       "      <th>zipcode</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ME</td>\n",
       "      <td>61756</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>NY</td>\n",
       "      <td>10017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>OH</td>\n",
       "      <td>30130</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  state zipcode\n",
       "0    ME   61756\n",
       "1    NY   10017\n",
       "2    OH   30130"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = pd.DataFrame({\n",
    "    \"state\":[df.iloc[0][\"state\"], \"NY\", \"OH\"],\n",
    "    \"zipcode\":[df.iloc[0][\"zipcode\"], \"10017\", \"30130\"],\n",
    "    \"type\":[\"in sample\", \"state in sample\\nzipcode out of sample\", \"out of sample\"]\n",
    "})\n",
    "\n",
    "X.iloc[:,:2]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d4dac1b8",
   "metadata": {},
   "source": [
    "We're making some quick convenience functions to map this new data to the proper indexes from the fitted model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "72a047fa",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Helper functions\n",
    "zip_lookup = dict(zip(df.zipcode.cat.categories, range(len(df.zipcode.cat.categories))))\n",
    "state_lookup = dict(zip(df.state.cat.categories, range(len(df.state.cat.categories))))\n",
    "\n",
    "def labels_to_index(series, lookup):\n",
    "    '''Converts categories to their proper codes'''\n",
    "    series = series.copy()\n",
    "    in_sample = series.isin(lookup.keys())\n",
    "    series.loc[~in_sample] = \"out_of_sample\"\n",
    "    return series.map(lookup).values.astype(\"int8\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "981d7f99",
   "metadata": {},
   "source": [
    "And finally we'll use the model to make predictions on this new data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "0ee0dc09",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='4000' class='' max='4000' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      100.00% [4000/4000 00:40<00:00]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "with model_nc:\n",
    "    # Set new data for the model to make predictions on\n",
    "    pm.set_data({\n",
    "        \"state_\": X.state.pipe(labels_to_index, state_lookup),\n",
    "        \"zip_\": X.zipcode.pipe(labels_to_index, zip_lookup)\n",
    "    })\n",
    "    \n",
    "    # make predictions\n",
    "    preds = pm.sample_posterior_predictive(trace_nc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "da3b9bf0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0, 'mu_zipcode')"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x453.6 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Rearrange predictions to make plot more readable\n",
    "pred_dct = {}\n",
    "for i in range(len(X)):\n",
    "    zipc = X.iloc[i][\"zipcode\"]\n",
    "    name = X.iloc[i][\"type\"]\n",
    "    pred_dct[f\"zip[{zipc}]\\n\"+name] = preds[\"y\"].T[i]\n",
    "\n",
    "# Plot predictions\n",
    "pm.plot_forest(pred_dct)\n",
    "plt.xlabel(\"mu_zipcode\", fontsize=14)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4be08fed",
   "metadata": {},
   "source": [
    "Notice how in sample predictions have smaller uncertainty intervals - this is great, its now much easier to make predictions compared to having to write out a custom prediction function that follows the same logic as the model"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}