data for estimation

[avi]

hi

I entered my model in a non linear form and I want to use dynare in order to log- linearized the model and estimate it (in a log-linear form).

I have 3 questions:

1) how should I write the model? should I use log-exp terms?

2) at the estimation function, should I use the log-linear command?

3) in what terms should i transform the data? should I use deviations from steady state (log and than deviation from hp filter), should I use log terms, and so on...

thanks

[StephaneAdjemian]

Hi,

I entered my model in a non linear form and I want to use dynare in order to log- linearized the model and estimate it (in a log-linear form).

I have 3 questions:

1) how should I write the model? should I use log-exp terms?

A much simpler approach is to add measurement equations in the model block of your mod file. For instance, you would write:

Code: Select all

GDPobs = 100*log(GDP/GDPss) ;


where GDP is defined in your non linear model and GDPss is a parameter for the steady state level of GDP.

2) at the estimation function, should I use the log-linear command?

If you follow my suggestion you don't need this option.

3) in what terms should i transform the data? should I use deviations from steady state (log and than deviation from hp filter), should I use log terms, and so on...

It's up to you. The one thing that matters is that the transformations on your data must be consistent with the declaration of the variables in the model.

Best, Stéphane.

[avi]

thank you for your answer. it was very useful