| Abstract: | According to Simpson's paradox (E. H. Simpson, 1951), if 2 or more contingency tables are collapsed into one, the resulting table may show a relationship between variables different from those shown by any of the original tables. Thus, a positive or negative relationship or stochastic independence may be shown by every component table, but be masked, in the collapsed table, by one of the other characteristics. This paradox has implications for the analysis of memory retrieval, particularly when the focus of interest is the relationship between success and failure on 2 retrieval attempts. Several recent issues in the memory literature are discussed in this connection, with emphasis on confounding effects of S differences, item differences, and S-item interactions. (23 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved) |
