John Spinelli, B.C. Cancer Research Centre
Goodness-of-Fit for Discrete Data
Federico O'Reilly, Universidad Nacional AutÃ³noma de MÃ©xico
Avoiding Asymptotics in Goodness-of-Fit
Disadvantages of needing different tables for testing the fit of different distributions, nowadays is diminished or bypassed by computer intensive procedures that calculate p-values. Some recent approaches followed in goodness-of-fit are suscintly mentioned and all of them shown to be related to an early result by Professor Durbin. Amongst the recent approaches, besides his well known work in the area, Professor Stephens has made several contributions along this more modern lines of research that span his contributions in goodness-of-fit for more than 50 years.
John Petkau, University of British Columbia
Evaluating Progression in Multiple Sclerosis Clinical Trials
Progression of multiple sclerosis (MS) is usually expressed in terms of the Extended Disability Status Scale (EDSS), originally developed by Kurtzke to quantify disability in MS. EDSS scores derive from a neurological examination in which multiple functional systems are evaluated. The EDSS is an ordinal scale, ranging from 0 (normal) to 10 (death) in half-points steps; scores up to 4.5 refer to fully ambulatory patients, while higher scores reflect increasing impairment to ambulation. The EDSS is often criticized for placing too much emphasis on ambulation and being relatively insensitive to clinical change, but it remains the standard for describing disability.
Progression of MS, as described by changes in EDSS, is typically the primary outcome measure for Phase III clinical trials in secondary progressive MS, and sometimes in relapsing-remitting MS. Repeated EDSS scores are collected on patients in Phase III MS trials; a typical schedule might be every 3 months for 3 years of follow-up. Despite the resulting richness of these primary outcome data sets, they are often analyzed in a very simple fashion.
I will review methods used for these analyses in MS clinical trials, starting with that originally recommended by Kurtzke and including those utilized in more recent trials that led to the identification of effective therapies for MS. I will also sketch some methods that can be used to make more effective use of the longitudinal EDSS data.
Ted Anderson, Stanford University
Likelihood Ratio Tests in Reduced Rank Regression and Blocks of Simultaneous Equations
When the rank of a a multivariate regression matrix is restricted, the matrix satisfies a set of linear restrictions. The rank of the restriction matrix is complementary to the rank of the regression matrix. Likelihood ratio tests of a specified regression matrix and of a specified restriction matrix are studied. Applications to econometric models are made.