Skip to main content
. 2020 Jun 2;9:e51260. doi: 10.7554/eLife.51260

Appendix 1—table 3. Test for an inverted-U relationship between choice behavior and DA baseline.

LooLM - LooQM ß2 estimate ß2p-value
sEBR WMC sEBR WMC sEBR WMC
model-based:
β −0.06 −0.04 −2.09e−04 2.98e−04 .132 .949
φ −3.60 −2.57 −1.13e−03 1.27e−02 .470 .809
ρ −53.09 −49.68 1.69e−03 1.20e−01 .869 .726
model-free:
payout −0.95 −1.05 −6.04e−04 1.37e−02 .582 .710
%bestbandit 198.06 −245.78 −2.45e−02 7.40e−02 .149 .897
meanrank 0.06 −0.10 −5.29e−04 −3.45e−03 .080 .733
%switches −484.04 −700.67 −2.09e−04 6.58e−01 .222 .509

Note. Choice behavior was assessed by the three choice parameters of the winning (Bayes-SM+E+P) model (upper part) and four model-free choice variables (lower part). Baseline dopamine (DA) function was assessed by the two behavioral DA proxies spontaneous eye blink rate (sEBR) and working memory capacity (WMC). For the latter, the first principal component across three different WMC tasks was used, denoted by WMCPCA. The column “LOOLM-LOOQM” denotes the difference of the squared distances for the linear model (LM) minus the quadratic model (QM) from the leave-one-out (LOO) model comparison. Note that negative values for LOOLM - LOOQM indicate better predictive accuracy of the LM. The columns “β2 estimate” and “β2 p-value” show for each quadratic model the estimated value and p-value of the β2 regression coefficient, respectively. Note that data from a pilot study (n=16) and the placebo condition of the main study were combined for this analysis to increase the sample size to n=47. β: softmax parameter; φ: exploration bonus parameter; ρ: perseveration bonus parameter.