Examples and Applications

Abstract

Three assumptions are necessary for meaningful inference under the mixture model. First, the predictor and criterion variables must be assumed to be bivariate normal so that the distribution of R, g(r) may be reasonably approximated by h(r). Second, the individual validity correlation coefficients must be independent. This key assumption is likely to be violated in some data sets as noted earlier (Section 3.3). A third assumption is that for each n _i sample size, there are the same t components in the conditional mixture with the same parameters λ _j and ρ _j .