Applied Mathematics Colloquium

Monday October 10, 2011 4:15 PM

Chi-Square and Classical Exact Tests Often Wildly Misreport Significance; the Remedy Lies in Computers

Speaker: Rachel Ward, Mathematics, University of Texas at Austin
Location: Annenberg 105
If a discrete probability distribution in a model being tested for goodness-of-fit is not close to uniform, forming the Pearson chi-square statistic often involves renormalizing summands to different scales in order to uniformize the asymptotic distribution. This often leads to serious trouble in practice -- even in the absence of round-off errors -- as the talk will illustrate via numerous examples. Fortunately with the now widespread availability of computers, avoiding all the trouble is simple and easy: without renormalization, the actual values taken by goodness-of-fit statistics are not humanly interpretable, but black-box computer programs can rapidly calculate their precise significance.

http://arxiv.org/abs/1108.4126

Joint work with Will Perkins and Mark Tygert.
Series Applied Mathematics Colloquium Series

Contact: Sydney Garstang at x4555 sydney@caltech.edu
For more information visit: http://www.acm.caltech.edu