topic GLS estimation in SAS Procedures

GLS estimation

alias — Mon, 26 Nov 2012 19:12:57 GMT

I've spent a long time looking for this so I thought I'd post for some help.

C1i = β1+β2Q1i+β3(Q1i)^2+β4(Q1i)^3+εi, Var(εi) = σ^2*Q1i

there are 28 obs. on Cost(C) and Output(Q).

Find the gls estimates for β1, β2, β3, and β4. I have been trying to find the code in SAS to do this but I've had no luck. I am trying PROC MIXED and PROC CALIS procedures in SAS. Does anyone know the code to answer this question?

The first column in the attached file is C1 and the second is Q1.

Thanks.

Re: GLS estimation

Reeza — Mon, 26 Nov 2012 19:52:11 GMT

Try proc GLM, see the example in the docs

http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_glm_sect048.htm

And some more tutorials here:

SAS Textbook Examples: Applied Regression Analysis, by Fox

Re: GLS estimation

Ksharp — Tue, 27 Nov 2012 03:03:48 GMT

Reeza,

I don't think GLM can do that .

Var(εi) = σ^2*Q1i which mean model has random effect ,GLM only can take care of Fix effect, therefore

PROC MIXED is a good choice. and SteveDenham may help you , you should post it at Statistical Forum, he is always there.

Ksharp

Re: GLS estimation

alias — Tue, 27 Nov 2012 10:12:45 GMT

Thanks for the help. I'm not very familiar with proc mixed or

proc calis, but that will change; when I have time. I've managed to get the correct answer using the just using the weight option in proc reg. I really would like to try the other options for the diagnostics they provide but I have a deadline.

Thanks again.