15 & 16 September 2016
Contemporary Issues in Hypothesis Testing
Often in Science, the causal bridge between the observable and the sought unknowns is obscure, if not an unknown itself. The relevance in such situations, of rejecting one hypothesis against another, stands challenged--as distinguished from the (Bayesian) model selection approach, in which the focus is on comparing different values of the unknowns given the available data. Recently, the classical hypothesis-significance-testing methods have been suggested as germane to contexts in which large, stable effects that stem from "small and well-controlled variation (the sorts of problems that Pearson, Fisher, etc., worked on)", but less so when the effects are "highly context dependent and with messy measurements". As expected, such pronouncement provoked reaction. The proposed workshop aims to provide a platform for the voicing of such opinions and reactions from statisticians and hands-on practitioners of the scientific method, particularly with the "messy"-ness of data unravelling upon us, as we proceed through the data revolution. Increased interaction amongst the community of hypothesis-testers/model-selectors is proposed, with keynote presentations by senior academics and talks by senior and junior colleagues, leading to discussions amongst all--with focus on the recent developments in the area, including the development of methods to test for the simplifying model assumption that helps overcome intractability, testing in high-dimensional models and/or data, nonparametric tests, fresh look at problems encountered in Bayes Factor computation, recent developments on multiple/composite testing, etc.
The workshop will take place on 15th and 16th of September, 2016 in MS.04 at the Department of Statistics (Zeeman building), Univeristy of Warwick. A detailed schedule can be found here.
Participants are encouraged to submit posters through the link. Posters will be accepted on a rolling basis with a final date of September 5, 2016 for poster submission.