variance values in model averaging

[gibbon]

I have a couple relatively simple question about model averaging. I have read through all the previous posts and while a few touch on the subject, I have not quite found the answer I need.

When averaging covariate estimates, which values from the variance-covariance matrix are fitted into the formula? Is there just one value or do all the covariance values for the covariate in question need to be included? Are the SE from the beta values not needed?

I apologize for these basic questions. Thanks in advance for any help.

Eric

[darryl]

Hi Eric,
You may also want to search the MARK forum, pretty sure the topic has been touched on before. Or, as it's a fairly general question on model averaging, research some of the more general literature on model averaging.
Cheers
Darryl

[gibbon]

Thanks for the reply! I found some useful answers in the MARK forum indicating that SEs can be changed into var by squaring them.
I now have one supplementary question:

I want to transform my betas in to real estimates before averaging. Do I need to transform the beta SEs in some way as well?

Thanks again.

Eric

[cooch]

Thanks for the reply! I found some useful answers in the MARK forum indicating that SEs can be changed into var by squaring them.
I now have one supplementary question:

I want to transform my betas in to real estimates before averaging. Do I need to transform the beta SEs in some way as well?

Thanks again.

Eric

All of that is covered at length in the MARK book.

[gibbon]

Thanks again for the reply. Today I finally got access to the B and A book and several things became tremendously more clear. This forum is also a great resource for 'new to the concepts AND the software' researchers like me.

Cheers, Eric