The RATS Software Forum

There is no set number. More is better than fewer.

you mention that the scale parameter and the degree of freedom need to be changed i order to model the inverse wishart distribution. In case I have degrees of freedom = number of endogenous variables +2 and and the prior mean of the main diagonal of the variance of the residuals matrix takes an inverse gamma distribution with scale and shape equal to .02, how do i model these into the code?

1. Why are you doing that?
2. It sounds like you're doing a hierarchical prior with an IW covariance matrix with diagonal scale matrix, with individually IG elements on the diagonal. I have no idea how you would do that. If that's not what you're intending, perhaps you should try writing down the formulas instead of explaining it in words.

In an earlier response you suggested that the simulated forecasts from Gibbsvar program can be compared with the point ahead forecasts with averaging. How many simulations do you suggest as appropriate for the averaging. Secondly you suggested that

For an informative Wishart prior, you need to add to RSSMAT the scale parameter for the I-Wishart prior and change the degrees of freedom in the %RANWISHARTI call.

compute rssmat=%sigmacmom(%cmom,bdraw)
*
* Draw sigma given b
*
compute sigmad=%ranwisharti(%decomp(inv(rssmat)),%nobs)

In case the scale parameter for the inverted wishart is .02 and the degrees of freedom are =the number of endogenous variables +2, how can these be incorporated in the rssmat and the ranwishart command
. Thanks in advance.

Please do not just start a new thread with a rehash of your previous questions.

Regarding the number of draws needed, I already answered that. That's the reality of working with simulation methods. Use too few and the simulation error tends to dominate. The simulation error goes down with the square root of the number of simulations.

The scale for an inverse Wishart isn't a number, it's a matrix.