This award to Northwestern University will allow the investigator to conduct methodological research on the effects of cluster sampling on the inferential procedures associated with log-linear analysis of cross-cultural survey data. The proposed research will examine the design effects of cluster sampling on 10-1 variables and on a large number of two- and three-way contingency tables for these variables. The design effects will be used to estimate the amount of bias in rejection rates. The study will be repeated on three cross-cultural samples varying in size, number of clusters, and geographical spread. Results from this analysis will be compared with the results obtained using a modification of a recently developed jacknife procedure that takes into account complex clustering and yuields robust and efficient inference. This research is important because cross cultural research is a major part of comparative research, and the methodological problems of comparing cultures across the world are significant. This project promises to advance our understanding of "Galton's problem", or the problem of how one can make inferences from cases where each case may not be totally independent of the others.
