“…As a general strategy to analyze our data, we opted for the Bayesian statistical analysis framework (for an introduction, see McElreath, 2020). Compared to null hypothesis significance testing, Bayesian analyses do not rely on p-values and statistical significance (Benjamin et al, 2018;Lima Portugal et al, 2020;McShane et al, 2019), but report P(θ | data), the probability distribution of the model's parameters (or quantities of interest derived from them) that are consistent with the model, observed data and prior information. Here, we summarize the uncertainty in our inference results by reporting the 95% credible intervals (95% CrI; 2.5%-97.5% quantiles) of the quantities of interest as well as the probability (P+ or P-) of the quantities of interest θ being greater or lower than 0, P+ = P(θ > 0 | data) or P-= P(θ < 0 | data).…”