The RATS Software Forum

I'm going through example 8.3 in the Switching Models Workbook and have a couple of queries:

1. The 3x1 vector NUPRIOR is filled with fours. In the workbook you write:

The second stage in this requires the degrees of freedom for the prior for each component (the NUPRIOR vector), which is 4 for all regimes in our example.

Could you please explain why there are 4 degrees of freedom in this example?

2. In my estimations of a two regime example, the elements of P tend to be (too) close to the priors I have chosen in the 2x2 gprior vector (even when ndraws is very large). Do you have any ideas about why this may be the case?

Thank you.

alexecon wrote:I'm going through example 8.3 in the Switching Models Workbook and have a couple of queries:

1. The 3x1 vector NUPRIOR is filled with fours. In the workbook you write:

The second stage in this requires the degrees of freedom for the prior for each component (the NUPRIOR vector), which is 4 for all regimes in our example.

Could you please explain why there are 4 degrees of freedom in this example?

It's a prior. 4 is small but not zero.

alexecon wrote: 2. In my estimations of a two regime example, the elements of P tend to be (too) close to the priors I have chosen in the 2x2 gprior vector (even when ndraws is very large). Do you have any ideas about why this may be the case?

That probably depends upon what you're using for GPRIOR. If the numbers in that are too large relative to the data set size, the prior will dominate the data.

This is from example 9.3 in the workbook:

gset MSRegime gstart gend = fix(%if(dx<0.0,1,2))

I understand that if dx is negative (0 or positive) it places a 1 (2) in series MSRegime but why is this necessary? The reason why I am asking is that with my data I have to change this (as I don't have any negative values).

That's just a guess as to what the regimes are. Whether what you put there even matters much depends upon how your guess values for the other parameters are set.