Merge branch 'main' of github.com:jt-lab/bambi into jt-main

bambinos · Oct 3, 2023 · b074ba1 · b074ba1
2 parents 9272a24 + 1387d44
commit b074ba1
Show file tree

Hide file tree

Showing 7 changed files with 3,288 additions and 3 deletions.
diff --git a/bambi/priors/scaler.py b/bambi/priors/scaler.py
@@ -114,7 +114,7 @@ def scale_threshold(self):
             threshold = self.model.components["threshold"]
             if isinstance(threshold, ConstantComponent) and threshold.prior.auto_scale:
                 response_level_n = len(np.unique(self.response_component.response_term.data))
-                mu = np.round(np.linspace(-2, 2, num=response_level_n - 1), 2)
+                mu = np.zeros(response_level_n - 1)
                 threshold.prior = Prior("Normal", mu=mu, sigma=1)
 
     def scale(self):

diff --git a/docs/_quarto.yml b/docs/_quarto.yml
@@ -65,6 +65,8 @@ website:
               - notebooks/circular_regression.ipynb
               - notebooks/quantile_regression.ipynb
               - notebooks/mister_p.ipynb
+              - notebooks/ordinal_regression.ipynb
+              - notebooks/zero_inflated_regression.ipynb
         - section: More advanced models
           contents:
             - notebooks/distributional_models.ipynb

diff --git a/docs/notebooks/data/hr_employee_attrition.tsv.txt b/docs/notebooks/data/hr_employee_attrition.tsv.txt
diff --git a/docs/notebooks/gallery.yml b/docs/notebooks/gallery.yml
@@ -87,7 +87,11 @@
     - title: Zero inflated models
       subtitle: When the outcome is mostly zeros and or is overdispersed
       href: zero_inflated_regression.ipynb
-      thumbnail: thumbnails/zero_inflated_pps.png
+      thumbnail: thumbnails/zip_model_pps.png
+    - title: Ordinal regression
+      subtitle: Model ordered category outcomes
+      href: ordinal_regression.ipynb
+      thumbnail: thumbnails/ordinal_regression.png
 - category: More advanced models
   description: ""
   tiles:

diff --git a/docs/notebooks/ordinal_regression.ipynb b/docs/notebooks/ordinal_regression.ipynb
diff --git a/docs/notebooks/thumbnails/ordinal_regression.png b/docs/notebooks/thumbnails/ordinal_regression.png
diff --git a/docs/notebooks/zero_inflated_regression.ipynb b/docs/notebooks/zero_inflated_regression.ipynb
@@ -1794,7 +1794,7 @@
     "\n",
     "In this notebook, two classes of models (ZIP and hurdle Poisson) for modeling zero-inflated data were presented and implemented in Bambi. The difference of the data generating process between the two models differ in how zeros are generated. The ZIP model uses a distribution that mixes two data generating processes. The first process generates zeros, and the second process uses a Poisson distribution to generate counts (of which some may be zero). The hurdle Poisson also uses two data generating processes, but doesn't \"mix\" them. A process is used for generating zeros such as a binary model for modeling whether the response variable is zero or not, and a second process for modeling the counts. These two proceses are independent of each other.\n",
     "\n",
-    "The datset used to demonstrate the two models had a large number of zeros. These zeros appeared because the group doesn't fish, or because they fished, but caught zero fish. Because zeros could be generated due to two different reasons, the ZIP model, which allows zeros to be generated from a mixture of processes, seems to be more appropriate for this datset."
+    "The dataset used to demonstrate the two models had a large number of zeros. These zeros appeared because the group doesn't fish, or because they fished, but caught zero fish. Because zeros could be generated due to two different reasons, the ZIP model, which allows zeros to be generated from a mixture of processes, seems to be more appropriate for this datset."
    ]
   },
   {
-Original file line number
+Diff line change
@@ Expand Up / @@ -1794,7 +1794,7 @@ @@
         "\n",
         "In this notebook, two classes of models (ZIP and hurdle Poisson) for modeling zero-inflated data were presented and implemented in Bambi. The difference of the data generating process between the two models differ in how zeros are generated. The ZIP model uses a distribution that mixes two data generating processes. The first process generates zeros, and the second process uses a Poisson distribution to generate counts (of which some may be zero). The hurdle Poisson also uses two data generating processes, but doesn't \"mix\" them. A process is used for generating zeros such as a binary model for modeling whether the response variable is zero or not, and a second process for modeling the counts. These two proceses are independent of each other.\n",
         "\n",
-        "The datset used to demonstrate the two models had a large number of zeros. These zeros appeared because the group doesn't fish, or because they fished, but caught zero fish. Because zeros could be generated due to two different reasons, the ZIP model, which allows zeros to be generated from a mixture of processes, seems to be more appropriate for this datset."
+        "The dataset used to demonstrate the two models had a large number of zeros. These zeros appeared because the group doesn't fish, or because they fished, but caught zero fish. Because zeros could be generated due to two different reasons, the ZIP model, which allows zeros to be generated from a mixture of processes, seems to be more appropriate for this datset."
        ]
       },
       {
@@ Expand Down @@