Estimation, shocks and observables

Dear all,

As far as I know, in case of estimation, number of shocks must be at least equal to the number of observable variables.

My question is that what if we have a deterministic model and want to estimate it, as in this kind of models we might have several shocks to the same variable during a period of time, would this number of shocks for the same variable be considered as just one shock or can be applied for the number of the observable variables as well ?!

I mean if let say we have 10 shocks to the same variable during the time can I take 10 observable variables or I should take just one?!

I know using a deterministic model even for the simulation purposes is not that common and it might get more complicated in case of estimation but I would appreciate any help.



Dynare estimates a model by maximizing the associated likelihood (or a weighted likelihood) function, which is the density of the sample conditional on the model and its parameters. Obviously if the model has no shocks it is not possible to define this density. I am not sure to understand correctly your idea. If your model defines an endogenous variable Y and you add two measurement errors E1 and E2, with variances s1 and s2 respectively, to this variable, then your model contains effectively only one shock with variance (with variance E1+E2). These two shocks are observationaly equivalent.


Dear St├ęphane,

Thanks for your precise reply. That is what I wanted to know.