Mix frequency data

Dear Prof. Pfeifer,

I ran a model with mixed frequency (in line with the example given in A Guide to Specifying Observation Equations for the Estimation of DSGE Models) to forecast the inflation, but the result seems dubious. Here’s files estim.mod (3.6 KB) data.csv (1.3 KB) . Kindly tell me what have I done wrong in this.



Please elaborate on what you mean with “suspicious”? What should I look for?

My goal is to forecast inflation rate (y_tilde) for next 12 months. I’m using year-on-year inflation rate as the observable variable Pi_tilde and filtered demeaned (intensive form) quarterly GDP as observable variable y_tilde_qtr. The problem is when I run the model forecasts are tend to centered around zero. Obviously, this seems wrong given my past inflation data is on average stays around 4-to-5. This may be because the steady state nominal inflation rate (Pibar) parameter is fixed at 1. But if I try to change this (Pibar), the steady state won’t be calculated. Your help and advice is much appreciated.

You need to define a proper observation equation. Your data was multiplied by 100 and is year on year. Your model variable is not multiplied by 100 and quarter on quarter. Also, you need to deal with the mean as well by adding a constant.