Explicit prior boundary problem

Svit

2026 4 20

0 个回答

44 次查看（30 天）

0 个投票

Model15.txt

I have a question about Dynare file, that I am trying to run in MATLAB.

For my master's thesis, I need to test certain economic parameters for the presence of structural breaks in parameters of DSGE model on the real world data for Slovenia, Austria and Italy for 1991-2020 (quarterly data). For the testing, I use so called "ESS procedure", developed by economists Atsushi Inoue and Barbara Rossi. The procedure includes 3 steps. The first step is about declaring "initial values" for the parameters. For most parameters, this works perfectly. But there is a problem with parameter gamma.

So, my problem is about explicit prior boundary of the parameter gamma (ie. interest rate elasticity of money demand), in other words, my problem is the non-positive definite Hessian at the mode. I ran into corner solutions and I am not sure what to do. AI suggests I fix the uniform prior (increase both mean and standard deviation) or use other distribution (either gamma_pdf or lognormal_pdf).

However, after increasing standard deviation from 0 to 10, there is no output like earlier, i get just 2 charts instead of 4 and much less text output.

So, is increasing standard deviation bad idea and, if yes, what should I do instead?

Thank you so much in advance!