
Dynare keeps including parameters which are not in the estimation anymore. I estimated once with all parameters and then excluded some from the estimated_params declaration. Then they keep appearing in the smoothed shocks and variance decomposition graphs. I tried clearing everything and deleting all the files created with the previous estimation but the problem persists. I guess this is still saved somewhere from the previous estimation, but how can I delete it?

What is the right interpretation of the historical variance decomposition of observables? In particular, does the direction of the shock show the sign of the shock itself or the sign of the effect that the shock has on the observable? For example: a positive bar of the MP shock on Consumption observable during recession. Does it mean that there has been a positive MP shock, or a negative MP shock which contributes positively to the Consumption observable?

What do you mean? The display of smoothed shocks has nothing to do with which parameters you estimate. If you have a shock in your model that has nonzero variance (e.g. because you set it in the shocksblock), it will appear.

It shows the shockâs contribution to the variable, not the shock itself. A positive bar of the MP shock on Consumption shows that the MP had a positive contribution to consumption, driving it up. Typically, this is a negative monetary policy shock that has such a positive effect on consumption.