Agenda

Sylvia Fruehwirth-Schnatter (Institute for Statistics and Mathematics, WU Vienna University of Economics and Business, Austria) - Skew Sparse Bayesian Factor Analysis

Abstract
Factor analysis is a popular method to obtain a sparse representation of the covariance matrix of multivariate observations and to uncover the unobserved driving factors behind observed correlation.  A challenge for factor models is to estimate the unknown number of factors and to recover an interpretable factor loading matrix from the data. Research in the area of sparse Bayesian factor analysis successfully addresses these issues within a Bayesian framework through the help of variable selection and shrinkage priors, see Frühwirth-Schnatter, Hosszejni and Lopes, 2025 for a recent review.
Most approaches to sparse Bayesian factor analysis, however, rely on the assumption that the idiosyncratic errors are Gaussian. In the present talk, this restrictive assumption is alleviated by assuming that the idiosyncratic error follow Azzalini’s multivariate skew-normal or skew-t distribution. Such a non-Gaussian factor model has a stochastic representation as a Gaussian factor model with an additional factor following a truncated standard normal distribution and the skewness parameters acting a factor loadings. This representation is exploited in the present talk to perform sparse Bayesian inference and learn the number of factors and the sparse loading matrix also for non-Gaussian factor models.
Applications to financial time series will serve as an illustration.

The seminar can be attended also remotely, connecting to ZOOM.
Zoom link
ID riunione: 837 8337 5271
Passcode: 6tRBL4