Fig. 5From: Explainable AI enables clinical trial patient selection to retrospectively improve treatment effects in schizophreniaComparison of actual post-treatment PANSS scores for the full sample and the treatment-indicated subgroup. Bars display treatment (TRT) and placebo (PBO) arms for the full randomized sample (left graph, TRT n = 95, PBO n = 102) with an illustrative instance of the BRL-classified treatment-indicated (TRT-indicated) subgroup (right graph, TRT n = 41, PBO n = 46). At a Cohen’s d of 1.24, the effect size between arms for the treatment-indicated subgroup is increased more than 50% over the effect size of the full sample (d = 0.82). Error bars are 95% confidence intervals on the meanBack to article page