dear jpfeifer,

i am using bayesian method to estimate the parameters in a log-linearized model

i have a data with, gdp, consumption, investement, prices, interest rate, real excange rate, current account to gdp ratio, inflation and foreign inflation.

the variables prices, interest rate, real excange rate, current account to gdp ratio, inflation are not stationnary.

so can i use the first difference for gdp, consumption, investement and one sided hp_filter for prices, interest rate, real excange rate, current account to gdp ratio, inflation.

and when i want to compare data moments with model’ moment, shoud i use the hp_filter option in the stoch_simul command.

thank you

Compare model and data moments is tricky in this case.The way you estimate your model, you are making your model explain the frequency bands in the data you use for estimation. Thus, the proper benchmark here would be to use the unfiltered model data originating in the respective observation equation and compare it to linked empirical data used in that observation equation. I.e. compare filtered interest rates from the data to the unfiltered interest rates in the model (which is the mapping the observation equation prescribes).

That being said, I do not understand why you use two different filters and filter the stationary data at all.

thank you professor for your reply

i a using filters for non stationnary variables (prices, interest rate, real excange rate, current account to gdp ratio, inflation) but for non stationnary variables (growth rate for gdp, investement and consumption) i do not use the filter.

thank you again

Taking first differences is a stationarity-inducing filter. And why do you think interest rates are not stationary?

in my data used for estimation, observables variables (prices, interest rate, real excange rate, current account to gdp ratio, inflation) are not stationnary, i’m used a unit_root test for these variable and i find that they are not stationnary, and their standard deviation are too high so i used the one sided hp_filter for these variables

and for the gdp, investement and consumption i used the first difference (ie: growth rate without hp_filter). is it correct professor?