Page 141 - Demo

Absence of cognitive bias to flanges in orang-utans1396This is reflected in the different interpretation of frequentist and Bayesian confidence intervals: The first is a range of values that contains the estimate in the long run, while the latter tells which parameter values are most credible based on the data (Kruschke et al., 2012; McElreath, 2018). Furthermore, Bayesian methods allow for the inclusion of prior expectations in the model, are less prone to Type I errors, and are more robust in small and noisy samples (Makowski et al., 2019). Altogether, these reasons make Bayesian methods a useful tool for data analysis. All models were created in the Stan computational framework and accessed using the brms package (Bürkner, 2017, 2018), version 2.18.5. All models were run with 4 chains and 6000 iterations, of which 1000 were warmup iterations. We checked model convergence by inspecting the trace plots, histograms of the posteriors, Gelman-Rubin diagnostics, and autocorrelation between iterations (Depaoli & van de Schoot, 2017). We found no divergences or excessive autocorrelation in any model. Furthermore we used the package emmeans (Lenth, 2023) to obtain posterior draws for contrasts. Below, we discuss the specific statistical models for each experiment.Dot-probe taskIn line with previous studies (Lacreuse et al., 2013; Laméris et al., 2022; Leinwand et al., 2022; van Berlo et al., 2023) we filtered the reaction times (RTs). First, we excluded slow reaction times, because they might reflect low motivation or distraction. Instead of opting for a fixed outlier criterion (e.g., Leinwand et al., 2022), we determined the upper limit per subject based on the median absolute deviation (MAD) in RT (i.e., RT = median + 2.5 × MAD; Leys et al., 2013). Second, we excluded reactions times <200ms, because they very likely represent anticipatory responses (Whelan, 2008). After the repetition of these unsuccessful trials, we applied the same filtering criteria. For the flange size dot-probe, we collected 423 trials of which 96 were excluded based on the outlier criteria (22.69%). In the repetition sessions, we collected 105 trials, 28 of which were excluded based on the outlier criteria (26.67%). Thus, our final dataset for the flange size dot-probe contained 404 trials (Kawan: 133; Samboja: 131; Sandy: 140). For the symmetry dot-probe, we followed the same procedure. In total, we collected 474 trials, 102 of which were excluded based on the outlier criteria (21.61%). In the repetition sessions, we collected 108 trials, 32 of which were excluded (29.63%). Thus, our final dataset for the symmetry dotprobe contained 448 trials (Kawan: 152; Samboja: 142; Sandy: 154).Tom Roth.indd 139 08-01-2024 10:41