Erratic estimation behavior

[klg123]

I am running my .mod file to estimate a small open economy DSGE model and have concerns that it is behaving strangely as it is run. I recently used the exact same code but in the first version had a parameter for targeting the nominal exchange rate in the monetary policy function. In this version, I have simply removed that feature and left everything else untouched, including the priors so that I can compare the two models. I am using mode_compute = 6 which should automatically tune scale for an acceptance rate of about 1/3, but the estimation seems to be behaving erratically as it runs. Would someone mind taking a look at this and offering insight into what is happening here? Much appreciated.

[jpfeifer]

What do you mean with erratic behavior? The described behavior suggests that you have trouble finding the mode.
The problems seem to come from the data treatment. Your model is log-linearized, but you enter the interest rate in levels. That is wrong, because the model implies they are mean zero while the data has a positive mean (it is the real interest rate plus inflation). Please see Pfeifer (2013): "A Guide to Specifying Observation Equations for the Estimation of DSGE Models" on my homepage for details. There, you will also find examples.

Additionally, I suggest you take a look at your data. Plot them. You will immediately see an extreme seasonal pattern in e.g. the growth rate. Hence, some of your data are not seasonally adjusted. You need to fix this before estimation.

[klg123]

Thank you for noticing this! Much appreciated.