MATURITY AND MOLT STATUS OF CRUSTACEA - DETERMINING CLASSIFICATION WITHOUT PRIOR CRITERIA

Existing methodology for estimation is reviewed for the situation wher e, a priori, the existence of two or more groups can be postulated but , in contrast with discriminant analysis, there is no sample in which the correct categories are known. Such mixture models are applied to t wo data sets related to the maturity and molt status of snow crab, Chi onoecetes opilio, namely (1) the classification of crab as morphometri cally mature or immature on the basis of the chela height - carapace w idth relationship and (2) the determination of the number of molts of crab, during a known period at liberty, from the relationship of size at release and size at. recapture of tagged animals. In the latter exa mple, a theoretical constraint is imposed that links the relationships at the different stages. The solution is obtained by ''nesting'' an i terative procedure within an EM algorithm. The method permits hypothes es concerning the number of groups to be tested, including the hypothe sis that the data come from a single homogenous group, and each indivi dual is assigned a probability of belonging to a group.