Ricardo Masini, Sao Paulo School of Economics-FGV

Bridging factor and sparse models
Date
Dec 1, 2020, 11:00 am12:30 pm
Event Description

Common factor model or some sort of sparsity assumption is widely used as a way to impose low-dimensional structure in a high-dimensional estimation problem. In this talk, we discuss the pros and cons of those options and combine both in a supervised learning methodology to efficiently explore all the information in high-dimensional datasets. The method is based on a very flexible linear panel data model with both observable and latent common factors as well as idiosyncratic structure which is supposed to have a weakly sparse cross-section dependence. Both factors and idiosyncratic terms may consist of dependent stochastic process. The methodology is divided into three steps. At each step, remaining cross-section dependence can be inferred by a new test for covariance structure in high-dimensions.

Recording