I in my Bayesian estimates of a medium scale new Keynesian model, shock decomposition charts do not change when I make pontual changes in one equation of the model or in the priors, and also when a split the sample in two. Is it expected to be this way?
More exactly: the changes are minimum, and not perceptible in the charts, even when coefficients change (which happens especially with the change of priors).
No, that is rather unusual. Large changes in the parameter values should also result in changes of the shock decomposition.