Title: Linear regression with time series regressors
Authors: Suhasini Subbarao - Texas A&M (United States) [presenting]
Abstract: In several diverse applications, from neuroscience to econometrics, it is of interest to model the influence regressors have on a response. In many of these applications, the regressors have a meaningful ordering (usually a time series), but the number of regressors is very large. Linear regression, where the number of regressors $n$ is of the same order or magnitudes larger than the observed number of responses $p$ has received considerable attention in recent years. However, most of these approaches place a sparsity assumption on the regressor coefficients. When the regressors are a time series, the sparse assumption can be unrealistic with no intuitive interpretation. We consider the problem of linear regression with stationary time series regressors, but work under the weaker assumption that the regressor coefficients are absolutely summable. We propose a computationally efficient method for consistently estimating the regression parameters, that avoids matrix inversion.