SGR Modeling of Correlational Effects in Fake Good Self-report Measures

摘要

In many self-report measures (i.e., personality survey items and diagnostic test items) the collected samples often include fake records. A case of particular interest in selfreport measures is the presence of caricature effects in participants’ responses under faking good motivation conditions. We say that a pattern of fake responses is a caricature pattern if it shows higher structural intercorrelations among faked items relative to the expected intercorrelations under the corresponding uncorrupted responses. In this paper we generalized a recent probabilistic perturbation procedure, called SGR - Sample Generation by Replacements - (Lombardi and Pastore (2012) Multivar Behav Res 47:519–546), to simulate caricature effects in fake good responses. To represent this particular faking behavior we proposed a novel extension of the SGR conditional replacement distribution which is based on a discrete version of the truncated multivariate normal distribution. We also applied the new procedure to real behavioral data on the role of perceived affective self-efficacy in social contexts and on self-report behaviors in reckless driving.