Fig. 4. Prior and Posterior distributions over β₁.

a: prior distribution over β₁. b: posterior distribution over β₁ along with median (red, dashed line), 95% Highest Density Interval (HDI-95) spanning [-0.03662-0.47061] and Region of Practical Equivalence (ROPE) at [-0.05, 0.05]. c: posterior distribution over expected lnRMSSD values as a function of symptoms' improvement for a subject recruited at the onset of a manic episode, identified with the dummy subject-id a. The posterior for the other subjects is available in Supplementary Fig. 2. Each black line (a total of one hundred is herewith displayed to avoid clutter) represents a single draw from the posterior, while the dashed green line is the average across all black lines sampled from the posterior. This illustrates how the Bayesian framework naturally incorporates uncertainty in its outputs, as in this plot we indeed have a distribution over lines and not just a single line. This notion of uncertainty enables better-informed decisions in a clinical setting, e.g. the confidence in a given positive trend in lnRMSSD is higher when lines are tightly packed around the average value.