Resumen de Comments on II: Statistical inference and large-scale multiple testing for high-dimensional regression models

Ya'acov Ritov

• We consider the estimation of a one-dimensional parameter in a linear model with an ultra-high number of independent variables. We argue that the standard assumptions on the design matrix are essentially technical and can be relaxed. Conversely, the assumptions on the sparsity of the nuisance parameters are unverifiable, too strong, and unavoidable.