Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: Chow test

From   "Kaganova, Yevgeniya" <>
To   <>
Subject   RE: st: Chow test
Date   Thu, 1 Apr 2010 09:08:51 -0400

Thank you very much! 

-----Original Message-----
From: [] On Behalf Of Clive Nicholas
Sent: Thursday, April 01, 2010 1:18 AM
Subject: Re: st: Chow test

Yevgeniya Kaganova wrote:

> Dear Statalisters, we want to justify that we can run a single model 
> with 3 groups of people in it (normal weight, overweight, and obese 
> people). We are generating coefficients on all vars for each group, 
> then just want to test our "disabled" variable across all the 3 groups 
> in the model post-estimation to see if different across the groups. 
> The testing for 3 groups is done the way it is explained in here 
> But the model we are running is GLM Y  X  if Y>0  family(gamma)
> link(log) . So the coefficients are logged. Is it OK to do is the same 
> way as it is done for the OLS. Can we test the 3 "disabled" 
> coefficients without transforming them first back to normal (unlogged) 
> scale. Do we have to worry about smearing factors with this GLM specification?

Happily assuming that -glm- is the way to estimate your model, what's wrong with combining those three groups to make an ordinally-measured index variable, and then running your model with that? In the interests of model parsimony - and being really lazy - that's what I would do.

Clive Nicholas

[Please DO NOT mail me personally here, but at <>. Please respond to contributions I make in a list thread here. Thanks!]

"My colleagues in the social sciences talk a great deal about methodology. I prefer to call it style." -- Freeman J. Dyson.

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index