A discussion of 'Sparse Bayesian factor analysis when the number of factors is unknown' by Sylvia Fruhwirth-Schnatter, Darjus Hosszejni and Hedibert Freitas Lopes

Hauzenberger, Niko and Koop, Gary (2024) A discussion of 'Sparse Bayesian factor analysis when the number of factors is unknown' by Sylvia Fruhwirth-Schnatter, Darjus Hosszejni and Hedibert Freitas Lopes. Bayesian Analysis. ISSN 1936-0975 (In Press) (https://doi.org/10.1214/24-BA1423)

[thumbnail of Hauzenberger-Koop-BA-2024-A-discussion-of-Sparse-Bayesian-factor-analysis-when-the-number-of-factors-is-unknown] Text. Filename: Hauzenberger-Koop-BA-2024-A-discussion-of-Sparse-Bayesian-factor-analysis-when-the-number-of-factors-is-unknown.pdf
Accepted Author Manuscript
Restricted to Repository staff only until 1 January 2099.
License: Strathprints license 1.0
Download (602kB) | Request a copy

Abstract

Conditional on knowing the number of factors, r, analysis in static and dynamic factor models is straightforward for the Bayesian. However, inference on r is challenging. A Bayesian could use marginal likelihoods to select the number of factors (see Geweke, 1996). But in the standard big data setups nowadays (which involve a large number of variables/measurements m), this is computationally cumbersome, requiring the estimation of a large set of models that vary in r (≤ m).

ORCID iDs

Hauzenberger, Niko ORCID logoORCID: https://orcid.org/0000-0002-2683-8421 and Koop, Gary ORCID logoORCID: https://orcid.org/0000-0002-6091-378X;
CORE (COnnecting REpositories)