Calculating the Fisher Information
Posted: Mon Nov 07, 2011 10:43 pmHi all,
I'm working with a small DSGE model and looking to calculate the Fisher information (i.e. negative expectation of the second derivative of the log likelihood) at a given parameter value. I'd prefer to avoid doing this by Monte-Carlo integration, but haven't been able to find tools to calculate the information directly in Dynare. Is this implemented in any version of Dynare/ is there code available for the purpose?
If not, does anyone have recommendations on what approach to use? I've seen papers (e.g. Iskrev (2008) and his references) on some of the steps needed to calculate the information, but Iskrev deals only with the problem of finding the information for the structural parameters once you already have the information for the state-space models, and it seems that there are a number of approaches to deriving the information for the state-space, all of which are quite involved.
Any help would be greatly appreciated.
Isa
I'm working with a small DSGE model and looking to calculate the Fisher information (i.e. negative expectation of the second derivative of the log likelihood) at a given parameter value. I'd prefer to avoid doing this by Monte-Carlo integration, but haven't been able to find tools to calculate the information directly in Dynare. Is this implemented in any version of Dynare/ is there code available for the purpose?
If not, does anyone have recommendations on what approach to use? I've seen papers (e.g. Iskrev (2008) and his references) on some of the steps needed to calculate the information, but Iskrev deals only with the problem of finding the information for the structural parameters once you already have the information for the state-space models, and it seems that there are a number of approaches to deriving the information for the state-space, all of which are quite involved.
Any help would be greatly appreciated.
Isa