Overview
Carlos M. Carvalho is a statistician at the University of Texas at Austin McCombs School of Business, specialising in Bayesian sparse factor modelling, variable selection, and financial applications. He is a leading contributor to the development of spike-and-slab factor models and the horseshoe prior for sparse regression.
Key Contributions / Features
- Co-developed the sparse factor analytic probit model with free factor covariance V that fosters sparsity in the loading matrix B (Hahn, Carvalho, Scott 2012)
- Key insight: allowing correlated factors (V ≠ I) reduces the number of non-zero loadings needed in B relative to orthogonal-factor models
- Prior work on sparse Bayesian factor models for genomics (Carvalho-Lucas-Wang-Nevins-West 2008)
Related