A probabilistic approach for evaluating the sensitivity to Fake Data in Structural Equation Modeling.

In this paper we address the issue of evaluating the sensitivity of goodness-of-fit indices in structural equation modeling when fake data are considered in three different factorial models with varying sample sizes (n= 50, 100 and 200). The sensitivity evaluation is carried out by means of a simulation procedure which combines a standard Monte Carlo approach and a new probabilistic version of a recent data perturbation procedure called Sample Generation by Replacements (SGR, Lombardi, Pastore and Nucci, 2004). Probabilistic SGR (PSGR) will be used to generate data perturbations based on three different models of faking: fake-uniform, fake-good (deception) and fake-bad (malingering). For each scenario of faking the performance of four very popular goodness-of-fit indices (two absolute indices: GFI, and AGFI; and two incremental indices: CFI and NNFI) will be evaluated.