ABSTRACT
Sometimes, we may have a set of similar contingency tables. For example, for large-scale clinical trials where a large number of patients are required, it is common to involve multiple medical centers to help with study recruitments so that the trials can be completed in a timely fashion. For example, one of the largest studies for treating alcohol dependence, COMBINE (Combined Pharmacotherapies and Behavioral Interventions), randomized 1,383 recently alcohol-abstinent subjects into 9 pharmacological and/or psychosocial treatment conditions from 11 academic sites in the United States (COMBINE Study Research Group, 2006). A study of this scale would have been much more difficult to conduct for a single medical facility. For rare diseases, such an approach is normally required because it is almost impossible to enroll enough patients at one site. Stratified studies also improve power; through stratification, subjects within the same stratum are more homogenous, and the reduced between-subject variability helps increase power. Since patients from different sites may be different in terms of their health conditions and varying levels of quality of health care services they receive from the different hospitals, treatments are likely to have varying effects across the sites. To account for such differences in the analysis, we cannot pull all patients’ data into one contingency table and apply the methods in Chapter 2.
