Predict-Then-Debias code

README.md

@@ -81,6 +81,7 @@ The package somes with a suite of examples on real data:
 - Plankton Counting with Computer Vision (label shift) ([```plankton.ipynb```](https://github.com/aangelopoulos/ppi_py/blob/main/examples/plankton.ipynb))
 - Ballot Counting with Computer Vision ([```ballots.ipynb```](https://github.com/aangelopoulos/ppi_py/blob/main/examples/ballots.ipynb))
 - Income Analysis with Boosting Trees ([```census_income.ipynb```](https://github.com/aangelopoulos/ppi_py/blob/main/examples/census_income.ipynb))
+- Tree Cover Analysis with Computer Vision (Predict-Then-Debias) ([```tree_cover_ptd.ipynb```](https://github.com/aangelopoulos/ppi_py/blob/main/examples/tree_cover_ptd.ipynb))
 
 # Usage and Documentation
 There is a common template that all PPI confidence intervals follow.
@@ -143,4 +144,6 @@ The repository currently implements the methods developed in the following paper
 
 [Prediction-Powered Bootstrap](https://arxiv.org/abs/2405.18379)
 
+[Prediction-Powered Inference with Imputed Covariates and Nonuniform Sampling](https://arxiv.org/abs/2501.18577)
+
 [The Mixed Subjects Design: Treating Large Language Models as Potentially Informative Observations](https://doi.org/10.1177/00491241251326865)

examples/README.md

@@ -17,4 +17,6 @@ Each notebook runs a simulation that forms a dataframe containing confidence int
 
 Each notebook also compares PPI and classical inference in terms of the number of labeled examples needed to reject a natural null hypothesis in the analyzed problem.
 
+The notebook [```tree_cover_ptd.ipynb```](https://github.com/aangelopoulos/ppi_py/blob/main/examples/tree_cover_ptd.ipynb) shows how to use the Predict-Then-Debias (PTD) estimator from Kluger et al. (2025), 'Prediction-Powered Inference with Imputed Covariates and Nonuniform Sampling,' https://arxiv.org/abs/2501.18577.
+
 Finally, there is a notebook that shows how to compute the optimal `n` and `N` given a cost constraint ([```power_analysis.ipynb```](https://github.com/aangelopoulos/ppi_py/blob/main/examples/power_analysis.ipynb)).

examples/power_analysis.ipynb

@@ -949,7 +949,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "base",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
    "name": "python3"
   },
@@ -963,7 +963,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.12.2"
+   "version": "3.9.7"
   }
  },
  "nbformat": 4,