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

st: Re: interaction terms in xtgee f(nb). interpretations incorrect???

To   "" <>
Subject   st: Re: interaction terms in xtgee f(nb). interpretations incorrect???
Date   Fri, 19 Jun 2009 08:30:35 +1000

Dear statalist,
further to my email below. Someone has recommended that I use dummy variables to create the interaction variables and have said that I can just use the coefficients for the purposes of my analysis. Sorry about my last post.

----- Original Message -----
From: <>
Date: Thursday, June 18, 2009 5:10 pm
Subject: interaction terms in xtgee f(nb). interpretations incorrect???
To: "" <>

> Dear stata list,
> I am currently using xtgee f(nb) corr(exch) to estimate a model 
> with interactions- and had a panic that I am interpreting them 
> incorrectly.I created two values value of the moderator, I std 
> dev above and below the mean. I then created an interaction with 
> each of these with the predictors in my model... (e.g. 
> family*1_stdabove and family*1_stdbelow). i tested whether these 
> were significant using the wald test and have so far just used 
> the coefficients of the model to interpret the results, not 
> predicted probabilities. 
> Any advice on whether this is ok? I know that there are some 
> program for nonlinear interactions such as spostado (by Long & 
> Freese) but I thought that this was used only if you want to 
> interpret predicted probs.
> Thanks in advance
> Allison
> ps- am currently waiting on the textbook "Regression Models for 
> Categorical Dependent Variables Using Stata, 2nd Edition"J. 
> Scott Long, Jeremy Freese but if anyone else has experience in 
> this I would be very grateful for their advice
*   For searches and help try:

© Copyright 1996–2022 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index