Sparse Dynamic Factor Models with Loading Selection by Variational Inference

TL;DR

This paper introduces a fast variational inference method for large, sparse dynamic factor models with loading selection, effectively handling missing data and enabling practical nowcasting applications.

Contribution

It develops a novel variational inference algorithm with slab-and-spike priors for sparse loading selection in dynamic factor models, suitable for real-time analysis.

Findings

Accurately identifies sparsity patterns in simulated data

Provides precise estimates of loadings and factors

Demonstrates computational efficiency for practical use

Abstract

In this paper we develop a novel approach for estimating large and sparse dynamic factor models using variational inference, also allowing for missing data. Inspired by Bayesian variable selection, we apply slab-and-spike priors onto the factor loadings to deal with sparsity. An algorithm is developed to find locally optimal mean field approximations of posterior distributions, which can be obtained computationally fast, making it suitable for nowcasting and frequently updated analyses in practice. We evaluate the method in two simulation experiments, which show well identified sparsity patterns and precise loading and factor estimation.