Closed gerreth closed 8 years ago
Dear Gerreth,
sorry for the slight delay. Forecasting is one of the most delicate features of RISE as many choices have to be made depending on the precise settings of your problem. Many questions have to be answered and they will influence the type of options you have to pass to RISE for it to behave the way you expect:
In any case, there are two functions for doing forecast in RISE. One is precisely the one you tried "forecast" and another one is "simulate". In fact, whatever you do from forecast goes through simulate. You will need, among other things,
The forecast horizon
run the command
help dsge/forecast
to see what your options are and let us know about the progress you make.
Cheers,
Junior
Hello,
no problem for the slight delay, I really appreciate it that you take the time to answer the questions here, besides providing this toolbox itself.
Regarding your questions to my problem:
I think I have problems with providing the database correctly. When I inspect the initial conditions of my variables (inside the @rise_generic/simulate file), the observables are set to their correct values (coming from the dataset used for the estimation). All other endogenous variables are 0. Is there a convenient way to set the initial conditions correctly? Changing the forecast horizon, start point and switching the shock uncertainty on/off works as intended as it seems.
Best Regards Gerret
The first thing to do then is to look at the historical database problem.
get(m,'endo_list(state)')
Unfortunately perhaps, many of those variables will be unobserved and if that the database you provide does not include data for those variables, RISE will use the steady state values. Actually is the other way around, i.e. RISE initializes all the variables in the state vector at their steady state values. It then uses the information provided in the database to overwrite these values.
2) One way to get initial conditions for the observable variables is to run the filter on the observables. In that way you get some initial conditions you can use.
Best, Junior
Dear Junior,
I managed to set the initial conditions in the case of a constant parameter model. However, in the case of a model with parameter switching I do get filtered variables for each regime. Which values do I set here as initial conditions?
Best Gerret
Hi Gerreth,
This is the type of choices I was hinting you would have to make. For the observed variables, the smoothed variables are the same. But for the other variables the values will most likely be different from one regime to another. You have two choices, depending on what you believe in:
Let us know if this helps,
Cheers,
Junior
This perfectly makes sense to me. I have somehow overlooked the expected values completely, but thought of doing the same thing. I will have to work on this, so many thanks for solving my problems up to now!
Dear Junior,
I have a question considering forecasting with an estimated dsge model. Simply doing 'forecast(model)' does not seem to work. What’s the right way to do this?
I guess I have to work on the usual forecasting problem first, but maybe already a follow up question, is it possible to use information on a mixed frequency when doing the forecast? For example, use information of an indicator (which determines the transition probabilities) which is available on a monthly basis to do the forecasting in an otherwise quarterly model?
Thank you in advance Gerret