Hi @olaf.klein
Thank you for providing the modified plotting function and this very clear post. There are some features here that definitly belong to the next version of the Bayesian module, which is still a couple of months away due to the recent release of v1.4. Particularly, I think the indexing range feature and the switch to a more easily distinguishable point-estimator marker (e.g. triangle) is very high on the priorities list.
I quite like your idea of offering more flexibility to the user for deciding what to put in the plots. Perhaps, a more UQLab way would be to allow the 'predDist' name-value pair in uq_display_uq_inversion, which produces the predictive plots, to accept an options structure myPredDistOpts instead of just a logical.
myPredDistOpts could then be filled with all kinds of options and combinations such as
myPredDistOpts.Range = [1:10, 200:210];
myPredDistOpts.Sample.Post = true;
myPredDistOpts.Sample.Prior= true;
myPredDistOpts.Sample.PostPred = true;
myPredDistOpts.Sample.PriorPred = false;
These would then be considered for the plot upon calling
uq_display(myBayesianAnalysis, 'predDist', myPredDistOpts)
What do you think?