Metadata-Version: 2.4
Name: midasverse-midas
Version: 0.3.0
Summary: Multiple Imputation using Denoising Autoencoders
Project-URL: Homepage, https://github.com/MIDASverse/MIDAS2
Project-URL: Issues, https://github.com/MIDASverse/MIDAS2/issues
Author-email: Thomas Robinson <t.robinson7@lse.ac.uk>, Ranjit Lall <ranjit.lall@politics.ox.ac.uk>
License-Expression: MIT
License-File: LICENSE.txt
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Programming Language :: Python :: 3
Requires-Python: >=3.9
Requires-Dist: numpy
Requires-Dist: pandas
Requires-Dist: scipy
Requires-Dist: statsmodels
Requires-Dist: torch
Description-Content-Type: text/markdown

# MIDAS2

Implementation of MIDAS in PyTorch. See [Lall and Robinson (2022)](https://doi.org/10.1017/pan.2020.49) for the original paper.

In addition to migrating to `torch`, this new version adds the following functionality:

* Models can be fit on `X` and used to impute on new data `X'`
* Automatic detection of column-types

## Example usage

MIDAS2 follows the sklearn API, with fit and transform methods of an imputer object.

```python
from midas2 import MIDAS

# Create a MIDAS model
mod = MIDAS()

# Fit the model to data
mod.fit(X, epochs = 10)

# Multiply impute missing data
X_imputed = mod.transform(m = 10)
```

## CHANGELOG

* Alpha release including combination rules function (26/06/2025)
* Renamed the main module from 'MIDAS2' to 'model' (19/12/2024)
* Restructured the package for easier install (19/12/2024)
