Abstract ABSTRACT: Linear model and retrospective process tracing methods of judgment modeling are compared in terms of predictive validity and convergence between measures of cue importance, in addition, the reliability of the linear model is examined. The experimental task required each of 31 subjects to provide either a buy or no-buy decision for 45 stocks (each described by six information cues). For each subject, a linear model was estimated and a retrospective process tracing model was generated by reference to post-experimental verbal reports. The results indicated that (1) each approach exhibited predictive validity, but the retrospective process tracings exhibited a superior ability to replicate the observed judgments, (2) linear model and retrospective process tracing measures of cue importance were related (exhibited convergent validity), and (3) the linear models were generally reliable.
Larcker et al. (Sat,) studied this question.