EZ: An Easy Way to Conduct a More Fine-Grained Analysis of Faked and Nonfaked Implicit Association Test (IAT) Data

Abstract

Although faking on the Implicit Association Test (IAT) is a relevant problem, it has not yet been considered for the traditional IAT effect (D measure). Research has suggested that diffusionmodel-based IAT effects may be useful as IAT v is related to the construct-related variance and IAT a and IAT t0 have both been assumed to provide indications of faking. Recent research used fast-dm to reanalyze nonfaked and faked IAT data under various faking conditions (faking low vs. faking high scores in a naïve vs. informed manner). The results showed that faking affected IAT v . However, there was an impact on IAT a when people knew how to fake and had to fake low scores. Thus, diffusion model analyses deliver additional information, but they are also very complex to perform. The diffusion tool EZ is easy to handle and very powerful, but researchers do not yet know whether IAT v , IAT a , and IAT t0 deliver similar information about the components in IAT results when they are obtained with EZ. Thus, we used EZ to reanalyze the data set described above. The results from fast-dm and EZ were comparable, but EZ had somewhat higher statistical power. IAT v was impacted by faking, thus replicating the finding that diffusion model analyses cannot yet be used to completely separate construct-and faking-specific variance from each other. However, replicating and extending the findings that were obtained with fast-dm, informed faking had an impact on IAT a and IAT t0 , which might both serve as indicators of faking. Thus, our results indicate that EZ as well as fast-dm is a powerful tool that can help researchers to interpret IAT results.

A diffusion model approach for understanding the impact of 17 interventions on the race Implicit Association Test

“…Only faking on the IAT (d = -0.13) and evaluative conditioning with the GNAT (d = -0.08) decreased IAT t 0 . Supporting earlier theory and evidence (Klauer et al, 2007;Röhner et al, 2013;Röhner & Thoss, 2018), participants in the faking condition took more time before and after categorization on compatible trials relative to incompatible trials compared to the control condition. Unexpectedly, participants who took the evaluative conditioning with the GNAT showed similar behavior (albeit to a lesser extent).…”

“…To conduct diffusion models, we used the Excel-based EZ software from http://www.ejwagenmakers.com/papers.html. The EZ diffusion model has already been applied successfully to IAT data (e.g., Röhner & Thoss, 2018) and has demonstrated high statistical power to detect effects even in small samples (e.g., Van Ravenzwaaij et al, 2016).…”

“…Computation of diffusion-model-based IAT effects. Using the parameters estimated in EZ diffusion model analyses, we computed IATv, IATa, and IAT t 0 (Röhner & Thoss, 2018).…”

The EZ Diffusion Model: An overview with derivation, software, and an application to the Same-Different task

“…It has fewer parameters, and consequently requires less information to fit empirical data than its more complex counterpart. Likewise, the authors report that many researchers do not use the diffusion model even where it is effective because it is too difficult to fit (Wagenmakers et al, 2007; also see Röhner & Thoss, 2018). Consequently, the EZ model was created to have a simplified diffusion model which still captures the crucial variables while being significantly simpler and faster to apply than the full model.…”

A tutorial on how to compute traditional IAT effects with {R}

“…Recent research has highlighted the idea that diffusion model analyses (Klauer, Voss, Schmitz, & Teige-Mocigemba, 2007;Röhner & Ewers, 2016b) or the quad model (Conrey, Sherman, Gawronski, Hugenberg, & Groom, 2005) are useful for decomposing the processes underlying the IAT effect in a more detailed manner. If you want to add diffusion-model-based IAT effects to the traditional IAT effects, the tutorial by Röhner and Ewers (2016a) as well as the paper by Röhner and Thoss (2018) can provide additional information and a guide for how to do this.…”