Agent learing through Kalman Filter

This forum is closed. You can read the posts but cannot write. We have migrated the forum to a new location where you will have to reset your password.
Forum rules
This forum is closed. You can read the posts but cannot write. We have migrated the forum to a new location (https://forum.dynare.org) where you will have to reset your password.

Agent learing through Kalman Filter

Postby onthetopo » Tue May 30, 2017 4:41 am

Hello,

I am curious if anybody can implement in dynare a model in which the agents learns the parameter of a hidden state-variable through Kalman fitler, like in http://www.sciencedirect.com/science/ar ... 8914000177

A Dynare example would be highly appreciated!

Basically, what I am thinking needs to be done in dynare is that agent will have an exogenous parameter (hidden-state-variable), that is updated period by period through some external loop. I have no idea if it is possible to pass through this parameter into dynare period-by-period, nor do I have any idea how Dynare can calcualte moments in such a setup.

If a learning example is not available, any kind of example, in which a parameter loop is passed into dynare [I guess we just create a new mod file each time, if there is no better method], is appreciated.
onthetopo
 
Posts: 10
Joined: Tue May 16, 2017 11:22 pm

Re: Agent learing through Kalman Filter

Postby jpfeifer » Thu Jun 01, 2017 4:23 pm

Patrick Hürtgen actually used Dynare for this paper. Unfortunately, I haven't yet found the time to prepare his codes for dissemination and posting them online.
------------
Johannes Pfeifer
University of Cologne
https://sites.google.com/site/pfeiferecon/
jpfeifer
 
Posts: 6940
Joined: Sun Feb 21, 2010 4:02 pm
Location: Cologne, Germany

Re: Agent learing through Kalman Filter

Postby onthetopo » Fri Jun 02, 2017 4:17 am

Quick question, is there a way to declare the shock variables as observables in the estimation()
procedure? Since in the learning model, the agent only knows innovation but doesn't know what are the true shocks.

You probably know what I'm thinking, I want to feed in innovations from empirical data into this learning model, with hard coded Kalman learning in the mod file, and estimate other parameters [steady state Kalman gain and so forth, and other utility parameters].
I understand the answer is very likely to be no.

jpfeifer wrote:Patrick Hürtgen actually used Dynare for this paper. Unfortunately, I haven't yet found the time to prepare his codes for dissemination and posting them online.
onthetopo
 
Posts: 10
Joined: Tue May 16, 2017 11:22 pm

Re: Agent learing through Kalman Filter

Postby jpfeifer » Sun Jun 04, 2017 3:15 pm

I don't know whether this is feasible, but you can try. You cannot define exogenous variables as observed. But you can define auxiliary endogenous variables that are identical to the shocks, e.g.
Code: Select all
var a_aux...;
varexo a ... ;
model;
a_aux=a;
end;
varobs a_aux ...;
------------
Johannes Pfeifer
University of Cologne
https://sites.google.com/site/pfeiferecon/
jpfeifer
 
Posts: 6940
Joined: Sun Feb 21, 2010 4:02 pm
Location: Cologne, Germany


Return to Dynare help

Who is online

Users browsing this forum: No registered users and 6 guests