Skip to main content
. 2019 Apr 30;8:e42512. doi: 10.7554/eLife.42512

Appendix 2—figure 10. Predicting image difficulty using image-based metrics.

Appendix 2—figure 10.

(A) Expected prediction improvement over a baseline model for models fit to the data from Experiment 1 (Appendix 2—figure 3), as estimated by the LOOIC (Vehtari et al., 2016). Values in deviance units (−2 * log likelihood; higher is better). Error bars show ±2 1 SE. Percentages are expected prediction improvement relative to the saturated model. (B) Same as A but for the data from Experiment 3 (Appendix 2—figure 6). (C) Odds of a success for a one SD increase in the image predictor for data from Experiment 1. Points show mean and 95% credible intervals on odds ratio (exponentiated logistic regression weight). (D) As for C for Experiment 3.