B0580
Title: Clustering and pairwise multiple comparison of spatial-temporal data based on L1 penalty
Authors: Chi Tim Ng - Hang Seng University of Hong Kong (Hong Kong) [presenting]
Woojoo Lee - Inha University (Korea, South)
Youngjo Lee - Seoul National University (Korea, South)
Abstract: The pairwise multiple comparison problem can be solved by maximizing a penalized likelihood. This does not only rule out the possibility of drawing contradictory conclusions, but also provides a general method of constructing multiple comparison procedures. Family-wise error rate can be controlled via the tuning parameter of the penalty function. In addition, selection consistency can further be achieved so that all hypotheses are concluded correctly with probability goes to one under certain conditions. Applications to spatial-temporal data in economics and genetics are discussed.
