In the case of estimation, Instead of using the data in level and loglinear option in the estimation command, I use data hpfiltered by myself. Doesn’t dynare treat the data as deviation from Steady State and use it accordingly or it still take some manipulation on the data before estimaytion?
For example, if I input the hpfilter residual of log(GDP) and log(consumption), named as y_obs and c_obs, can I just write two more equations in the dynare model:
where err1 and err2 stand for some measurement error, and y and c are the theoretical variables in level in the model.
The rule is always to have variables with identical definition in the data and in the model. If your data are detrended before, the variables in the model must deviation from the steady state. This mean that you have to rewrite your model in deviation form.
Thanks, Michel. It is clear. I have one more question. If some of my variables needs log-linearization while others don’t, how can I do that in dynare or should I do it before throwing them in Dynare?
The best way to handle it is to use Dynare without the ‘loglinear’ option and consider only straight linearization. If you want loglinearization for one variable, express this variable in log as in example2.mod and use exp(x) when you need the level of x in one equation.