Page 1 of 1

where is contemporaneous cross-correlation?

PostPosted: Thu Jun 25, 2015 10:21 am
by jkh
I can't find contemporaneous cross-correlations between endo-variables after running bayesian estimation.

"oo_.PosteriorTheoreticalMoments.dsge.correlation" shows only lagged cross-correlations.

Where can I find them ?


Please help me!!!

Re: where is contemporaneous cross-correlation?

PostPosted: Thu Jun 25, 2015 5:39 pm
by jpfeifer
You need to construct them. Take a look at the GarciaCiccoetal2010.mod at https://sites.google.com/site/pfeiferecon/dynare

Re: where is contemporaneous cross-correlation?

PostPosted: Fri Jun 26, 2015 4:52 am
by jkh
Thank you for your quick reply.

But, this way doesn't seem to give confidence intervals for cross autocorrelations (ie. HPD region for Bayesian estimation)

Maybe, it seems to result from the single estimates for policy function.

However, I think that in fact, there exist distributions for policy function's parameter estimates.
As a result, cross-autocorrelation also should have confidence intervals.
Am I right?

In addtion, another related question is whether policy function's parameters in "oo_.dr.ys" come from Bayesian estimation's mode value or anything.

Thanks a lot!!!

Re: where is contemporaneous cross-correlation?

PostPosted: Wed Jul 01, 2015 5:33 am
by jpfeifer
Unfortunately, that is true. Currently, the distribution of contemporaneous correlations is not provided. You would need to construct them in each iteration. Code for doing this would be here: https://github.com/DynareTeam/dynare/pull/871

The values in oo_.dr.ys after Bayesian estimation come from the last subdraw. They are not at the mode. If you want this, you need to run stoch_simul once again.