Yearly data for bayesian estimation

Dear all:
I’m a beginner in DSGE, now i want to try use bayesian method to estimate some parameters, and the data i use is yearly data, this means the data in the mat file is yearly data, but i wander how the dynare know the data is quarterly data or yearly data, in other words, in which way can i tell the dynare in the mat file is the yearly data, i know for some parameters ,the frequency of data may change the value of parameters. I have read the dynare manual, but still can’t figure out.
Looking forward your help,thank you very much !

For Dynare the frequency of your data does not matter. A period is a period, regardless of whether a quarter or a year. It is the job of the model builder to make sure the parameterization matches the data frequency.

Thank you very much, this really helps me a lot!

What if my data is mixed with quarterly data and yearly data? In order to make it consistent, is it possible to transform yearly data to quarterly data, e.g. each quarterly data=yearly/4?


It is possible to transform yearly data into quarterly data, or quarterly data into monthly data, using some ad hoc time series tools. But these transformations are not model consistent and you may end up with undesired properties in the data. My preferred approach would be to write the model in the highest frequency and add measurement equations for linking, say, the theoretical monthly variables to, say, quarterly observed variables. For instance, for linking the monthly theoretical output flow to the quarterly observed output you would have something like:

Yobs = y + y(-1) +y(-2);

Because Yobs is not observed in each period (since some variables, say consumption, are observed monthly) you will have to use missing variables. Quarterly output would be observed in March, June, September and December. Other months should be treated as missing observations. (with NaNs in your database).


I see. Thanks a lot for your reply.

Section 7 on “Time Aggregation” in Pfeifer(2013): “A Guide to Specifying Observation Equations for the Estimation of DSGE Models” provides more details on this.

Thank you, Professor.
Best regards,