A1180
Title: Hierarchical Lasso for multi-factor designs
Authors: Tatyana Krivobokova - University of Vienna (Austria) [presenting]
Razvan Andrei Morariu - University of Vienna (Austria)
Abstract: In modern experimental sciences, high-throughput experimentation (HTE) generates large datasets that possess a very specific structure. For example, in chemical HTE, reaction properties are measured for all combinations of multiple components, forming a factorial design with a single observation per factor-level combination. A model reflecting this data-generating process should be estimated using a hierarchical Lasso estimator that enforces sparsity and hierarchical constraints on main effects and interactions. An efficient estimation algorithm is developed, and it is shown that the algorithm achieves consistent variable selection and estimation under mild conditions.
