Bayesian sparse factor models have proven useful for characterizing dependence in multivariate data, but scaling computations to large number of samples and dimensions is problematic. We propose expandable factor analysis for scalable estimation in factor models. The method relies on a novel multiscale generalized double Pareto shrinkage prior that allows efficient estimation of low-rank and sparse loadings matrices through weighted L1-regularized regression. Efficient sampling and estimation algorithms are developed that accommodate uncertainty in the number of factors and that generalize to other settings. The methods are applied to simulated data and genomic studies.
Joint work with Barbara E. Engelhardt (Princeton University) and David B. Dunson (Duke University)
|