+ ran precommit on new NB, created myst file

Author: jonsedar

examples/generalized_linear_models/GLM-missing-values-in-covariates.ipynb

@@ -146,6 +146,9 @@
    },
    "outputs": [],
    "source": [
+    "# suppress seaborn, it's far too chatty\n",
+    "import warnings  # #noqa\n",
+    "\n",
     "from copy import deepcopy\n",
     "\n",
     "import arviz as az\n",
@@ -157,9 +160,7 @@
     "\n",
     "from pymc.testing import assert_no_rvs\n",
     "\n",
-    "# suppress seaborn, it's far too chatty\n",
-    "import warnings  # #noqa\n",
-    "warnings.simplefilter(action=\"ignore\", category=FutureWarning) # noqa\n",
+    "warnings.simplefilter(action=\"ignore\", category=FutureWarning)  # noqa\n",
     "import seaborn as sns"
    ]
   },
@@ -859,6 +860,7 @@
     "    _ = f.suptitle(\"Univariate numerics with NaN count noted\")\n",
     "    _ = f.tight_layout()\n",
     "\n",
+    "\n",
     "plot_univariate_violin(df, fts=[\"y\"])"
    ]
   },
@@ -952,7 +954,7 @@
     "    fit_reg=True,\n",
     "    height=4,\n",
     "    aspect=0.75,\n",
-    "    facet_kws={'sharex':False}\n",
+    "    facet_kws={\"sharex\": False},\n",
     ")\n",
     "_ = g.fig.suptitle(\"Bivariate plots of `y` vs fts `a`, `b`, `c`, `d`\")\n",
     "_ = g.fig.tight_layout()"
@@ -2131,9 +2133,9 @@
     "FTS_XJ = [\"intercept\", \"a\", \"b\", \"c\", \"d\"]\n",
     "\n",
     "COORDS = dict(\n",
-    "            xj_nm=FTS_XJ,  # these are the names of the features\n",
-    "            oid=dfrawx_train.index.values  # these are the observation_ids\n",
-    "            )\n",
+    "    xj_nm=FTS_XJ,  # these are the names of the features\n",
+    "    oid=dfrawx_train.index.values,  # these are the observation_ids\n",
+    ")\n",
     "\n",
     "with pm.Model(coords=COORDS) as mdl0:\n",
     "    # 0. create (Mutable)Data containers for obs (Y, X)\n",
@@ -2344,7 +2346,7 @@
     }
    ],
    "source": [
-    "GRP = 'prior'\n",
+    "GRP = \"prior\"\n",
     "kws = dict(samples=2000, return_inferencedata=True, random_seed=42)\n",
     "with mdl0:\n",
     "    id0 = pm.sample_prior_predictive(var_names=RVS_PPC + RVS_PRIOR, **kws)"
@@ -2392,6 +2394,7 @@
     "    _ = f.suptitle(f\"In-sample {grp.title()} PPC Retrodictive KDE on `{ynm}` - `{mdlnm}`\")\n",
     "    return f\n",
     "\n",
+    "\n",
     "f = plot_ppc_retrodictive(id0, grp=GRP, rvs=[\"yhat\"], mdlnm=\"mdl0\", ynm=\"y\")"
    ]
   },
@@ -2562,7 +2565,7 @@
     }
    ],
    "source": [
-    "GRP = 'posterior'\n",
+    "GRP = \"posterior\"\n",
     "with mdl0:\n",
     "    id0.extend(pm.sample(**SAMPLE_KWS), join=\"right\")\n",
     "    id0.extend(\n",
@@ -2972,6 +2975,7 @@
     "    _ = f.tight_layout()\n",
     "    return f\n",
     "\n",
+    "\n",
     "f = plot_forest(id0, grp=GRP, rvs=[\"beta_j\"], mdlnm=\"mdl0\")"
    ]
   },
@@ -3371,21 +3375,23 @@
     "FTS_XJ = [\"intercept\", \"a\", \"b\"]\n",
     "FTS_XK = [\"c\", \"d\"]\n",
     "COORDS = dict(\n",
-    "            xj_nm=FTS_XJ,  # names of the features j\n",
-    "            xk_nm=FTS_XK,  # names of the features k\n",
-    "            oid=dfx_train.index.values  # these are the observation_ids\n",
-    "            )\n",
+    "    xj_nm=FTS_XJ,  # names of the features j\n",
+    "    xk_nm=FTS_XK,  # names of the features k\n",
+    "    oid=dfx_train.index.values,  # these are the observation_ids\n",
+    ")\n",
     "\n",
     "with pm.Model(coords=COORDS) as mdla:\n",
     "    # 0. create (Mutable)Data containers for obs (Y, X)\n",
     "    y = pm.Data(\"y\", dfx_train[ft_y].values, dims=\"oid\")\n",
     "    xj = pm.Data(\"xj\", dfx_train[FTS_XJ].values, dims=(\"oid\", \"xj_nm\"))\n",
     "\n",
     "    # 1. create auto-imputing likelihood for missing data values\n",
-    "    # NOTE: there's no way to put a nan-containing array (nor a np.masked_array) \n",
+    "    # NOTE: there's no way to put a nan-containing array (nor a np.masked_array)\n",
     "    # into a pm.Data, so dfx_train[FTS_XK].values has to go in directly\n",
     "    xk_mu = pm.Normal(\"xk_mu\", mu=0.0, sigma=1, dims=\"xk_nm\")\n",
-    "    xk = pm.Normal(\"xk\", mu=xk_mu, sigma=1.0, observed=dfx_train[FTS_XK].values, dims=(\"oid\", \"xk_nm\"))\n",
+    "    xk = pm.Normal(\n",
+    "        \"xk\", mu=xk_mu, sigma=1.0, observed=dfx_train[FTS_XK].values, dims=(\"oid\", \"xk_nm\")\n",
+    "    )\n",
     "\n",
     "    # 2. define priors for contiguous and auto-imputed data\n",
     "    b_s = pm.Gamma(\"beta_sigma\", alpha=10, beta=10)  # E ~ 1\n",
@@ -3709,7 +3715,7 @@
     }
    ],
    "source": [
-    "GRP = 'prior'\n",
+    "GRP = \"prior\"\n",
     "kws = dict(samples=2000, return_inferencedata=True, random_seed=42)\n",
     "with mdla:\n",
     "    ida = pm.sample_prior_predictive(\n",
@@ -3976,7 +3982,7 @@
     }
    ],
    "source": [
-    "GRP = 'posterior'\n",
+    "GRP = \"posterior\"\n",
     "with mdla:\n",
     "    ida.extend(pm.sample(**SAMPLE_KWS), join=\"right\")\n",
     "    ida.extend(\n",
@@ -5228,10 +5234,12 @@
     "\n",
     "    # same code as above for mdla\n",
     "    # 1. create auto-imputing likelihood for missing data values\n",
-    "    # NOTE: there's no way to put a nan-containing array (nor a np.masked_array) \n",
+    "    # NOTE: there's no way to put a nan-containing array (nor a np.masked_array)\n",
     "    # into a pm.Data, so dfx_holdout[FTS_XK].values has to go in directly\n",
     "    xk_mu = pm.Normal(\"xk_mu\", mu=0.0, sigma=1, dims=\"xk_nm\")\n",
-    "    xk = pm.Normal(\"xk\", mu=xk_mu, sigma=1.0, observed=dfx_holdout[FTS_XK].values, dims=(\"oid\", \"xk_nm\"))\n",
+    "    xk = pm.Normal(\n",
+    "        \"xk\", mu=xk_mu, sigma=1.0, observed=dfx_holdout[FTS_XK].values, dims=(\"oid\", \"xk_nm\")\n",
+    "    )\n",
     "\n",
     "    # 2. define priors for contiguous and auto-imputed data\n",
     "    b_s = pm.Gamma(\"beta_sigma\", alpha=10, beta=10)  # E ~ 1\n",