ON DETECTING NONLINEAR NONCOMPENSATORY JUDGMENT STRATEGIES - COMPARISON OF ALTERNATIVE REGRESSION-MODELS

We compare the performance of the two models that are usually used to detect nonlinear noncompensatory (NLNC) judgment strategies-Einhorn's (1970) parabolic and hyperbolic models-to two new models: (1) The scat ter model (Brannick and Brannick, 1989), which includes, in addition t o the linear combination of attribute values, a nonlinear term-the wit hin profile scatter of the attributes; and (2) The ''true conjunctive- disjunctive'' model (TCD model), which includes additional nonlinear t erms associated with the relative values of the attributes, The analys es of 12 empirical datasets indicate that the scatter model is the bes t NLNC model in terms of model fit, Furthermore, the analyses also ind icate that when judgment strategies are relatively homogeneous, the no nlinear terms of the scatter model and the TCD model are most useful m easures of NLNC strategies because they are statistically more powerfu l than model fit-based measures and because they permit the representa tion of judgment strategies on a conjunctive-disjunctive continuum, Th e nonlinear term of the scatter model also allows testing for NLNC str ategies on the individual level. (C) 1995 Academic Press, Inc.