Figure 3. Posterior model probabilities given the observed data (N = 62) and relationship between the attraction weight and the placebo effect.

(A) Single subjects posterior model probabilities for the Bayesian and Null model. The data is sorted by the Bayesian model posterior probability of each subject. (B) Random effects overall posterior model probability for the two models. (C) Positive correlation between the placebo effect and the Bayesian integrated treatment variability (attraction weight – relative variability of prior and likelihood, see Materials and methods Equation (3)). This is implying that higher treatment precision (prior) compared to higher variability in new sensory inputs (likelihood) may lead to larger placebo effects. See also Figure 3—figure supplements 1 and 2 for single subject fits. RFX, random effects; r, correlation coefficient.

Figure 3—figure supplement 1. Integration of prior and likelihood of single subjects including predictions for the Bayesian and Null model: high treatment precision subjects (HTP group).

Solid lines represent observed data and dashed lines represent model predictions of the placebo treatment outcome. pµ – prior µ; pσ – prior standard deviation; prob – posterior probability of the Bayesian integration model (compared to the no-learn Null model); pdf – probability density function; VAS – visual analogue scale.

Figure 3—figure supplement 2. Integration of prior and likelihood of single subjects including predictions for the Bayesian and Null model: low treatment precision subjects (LTP group).

Solid lines represent observed data and dashed lines represent model predictions of the placebo treatment outcome. pµ – prior µ; pσ – prior standard deviation; prob – posterior probability of the Bayesian integration model (compared to the no-learn Null model); pdf – probability density function; VAS – visual analogue scale.