Table 2.

MPH models. Median [25–75 percentile] of subject-level parameter estimates in model space. Absolute WAIC is reported as the estimate of model evidence, where a smaller WAIC indicates higher evidence. Biased instrumental learning rate for rewarded Go and punished NoGo responses as computed by ε₀±κ under placebo and by ε₀±(κ+κ_MPH) under MPH. (MPH) indicates the value of that parameter under MPH.

DOI: http://dx.doi.org/10.7554/eLife.22169.018

	Extended MPH models
	M5a	M5b	M5c	M6
WAIC	66383	66883	66595	66069
ρ	31.2 [14.7 53.6]	31.6 [15.6 57.0]	55.8 [19.6 104.8]	51.9 [20.6 98.7]
ε0	0.022 [0.010 0.067]	0.021 [0.011 0.061]	0.011 [0.006 0.051]	0.012 [0.006 0.055]
b	−0.04 [−0.33 0.18]	−0.05 [−0.34]	−0.10 [−0.37 0.13]	−0.14 [−0.42 0.10]
π π (MPH)	0.27 [−0.50. 71] 0.20 [−0.38. 71]	0.15 [−0.28. 70]	0.05 [−0.46. 61]	0.27 [−0.47. 74] −0.05 [−0.70. 50]
ε rewarded Go ε rewarded Go (MPH)	0.037 [.017. 116]	0.030 [.018. 103] 0.031 [.016. 104]	0.018 [.009. 082]	0.019 [.009. 085]
ε punished NoGo ε punished NoGo (MPH)	0.009 [.004. 030]	0.009 [.003. 021] 0.008 [.002. 021]	0.004 [.002. 013]	0.005 [.002. 017]
ε diffuse (MPH)			0.002 [.002. 004]	0.003 [.002. 004]