Extended Data Fig. 10: The reduced accuracy of the sc- models is consistent across noise levels and trial types.
From: Modelling human behaviour in cognitive tasks with latent dynamical systems
a, Mean error rate for congruent trials. b, Mean error rate for incongruent trials. c, Mean error rate for stay trials. d, Mean error rate for switch trials. e, Correlation between model robustness and participant congruency effect (center line shows the Pearson’s r at each noise level, error bars show bootstrap 95% CIs). Model robustness was defined as the difference in model accuracy between a given noise level (x-axis) and baseline (0.1SD). The 95% CIs did not contain zero for noise values ≥0.4SD. For panels a-e, error bars show bootstrap 95% CIs. Panels a-d: N = 25 sc+ models and 25 sc- models; panel e: N = 140 participants/models.
