In biomedical fields (and in most scientific fields I suppose), the majority of the data are time series. However, most of the statistical methods used reduce time series into a single, arbitrary, data point (such as mean, median or maximum). To avoid this considerable loss of information, some techniques exist (such as spm1d), but they are not well known and are based in the frequentist framework.
I have already successfully applied Bayesian models for scalar datasets, however I don’t really know how to formulate a time related model.
Consider the following example: participants from two populations (A and B) perform six tests of an experimental task generating three variables (x, y and z) of size 100 each (say 100 seconds). You want to obtain posterior distributions that consider the time component and co-variances of the variables to make inferences (e.g. how population A differs from population B). How would you formulate such model?