ahh of course, makes sense. follow up question - Indeed, it is evident that obtaining the posterior distribution for each variable and then extracting the mode by employing, for instance, the Scipy.stats.mode function, will not yield the maximum a posteriori estimate for this parameter. This aspect is also connected to my second assertion regarding predictive power and accounts for my reduced predictive ability when relying on mode-estimates of the parameters via the aforemented approach. I wonder if any relevant functions are accessible for obtaining the maximum a posteriori estimate from a trace, or if Arviz has conducted any investigations in this direction. Essentially, my aim is to construct a highest density interval around the maximum a posteriori estimate, with appropriate guardrails set at, for instance, 45%-55%. Consequently, I need to acquire the complete posterior distribution but restrict my attention to the 45-55% interval for each parameter surrounding the mode.