A Bayesian Approach to Spherical Factor Analysis for Binary Data

TL;DR

This paper introduces a Bayesian spherical factor analysis model for binary data, offering a flexible, interpretable geometric embedding approach that includes traditional models as special cases and allows for uncertainty quantification.

Contribution

It proposes a novel spherical manifold embedding model for binary data that generalizes traditional factor models and enhances interpretability and uncertainty quantification.

Findings

Model includes traditional factor models as a special case.

Demonstrates effectiveness through simulations and real voting data.

Provides a flexible and interpretable geometric embedding for binary data.

Abstract

Factor models are widely used across diverse areas of application for purposes that include dimensionality reduction, covariance estimation, and feature engineering. Traditional factor models can be seen as an instance of linear embedding methods that project multivariate observations onto a lower dimensional Euclidean latent space. This paper discusses a new class of geometric embedding models for multivariate binary data in which the embedding space correspond to a spherical manifold, with potentially unknown dimension. The resulting models include traditional factor models as a special case, but provide additional flexibility. Furthermore, unlike other techniques for geometric embedding, the models are easy to interpret, and the uncertainty associated with the latent features can be properly quantified. These advantages are illustrated using both simulation studies and real data on voting records from the U.S. Senate.