Re: st: Testing interaction terms

[Verena Dill <dill@uni-trier.de>]

What you wrote is exactly what I did in my regression (output below, 
just for the matter of illustration I included the four categories;  
var11: var1==1 & var2==1, var10: var1==1 & var2==0, var01: var1==0 & 
var2==1, var00: var1==0 & var2==0; because of the below mentioned 
structure of the data two of the categories are omitted).
------------------------------------------------------------------------------
     partner |      Coef.   Std. Err.      z    P>|z|     [95% Conf. 
Interval]
-------------+----------------------------------------------------------------
       var11 |   .7478253   .3528458     2.12   0.034     .0562602     
1.43939
       var10 |   .9636673   .3315029     2.91   0.004     .3139335    
1.613401
       var01 |          0  (omitted)
       var00 |          0  (omitted)
       _cons |    -.63364   .3095231    -2.05   0.041    -1.240294   
-.0269858
But my question is: how can I test if the coefficients _b[var11] and 
_b[var10] are equal taking the "interaction"-nature of the variables 
into account? Using only "test _b[var11]= _b[var10 ]"  does not account 
for that. Is there any other procedure I could use here (maybe similar 
to contrast)?


Am 19.06.2013 14:25, schrieb David Hoaglin:
> Verena,
>
> Because the data have no observations for var1==0&  var2==0, it is not
> possible to express the combined effect of those variables (in the
> linear predictor) in the usual way,
> (effect of var1) + (effect of var2) + (interaction).
> One alternative approach is to treat the combination of var1 and var2
> as a categorical variable with three categories: var1==0&  var2==1,
> var1==1&  var2==0, and var1==1&  var2==1.
>
> David Hoaglin
>
> On Wed, Jun 19, 2013 at 8:06 AM, Verena Dill<dill@uni-trier.de>  wrote:
> > I obtained the coefficients from a regression model and want to test whether
> > or not the coefficients are significantly different from each other.
> > The problem now is that the two variables are related to each other like
> > interactions and only partly overlap.
> >
> > var1: variable 1 (dummy)
> > var2: variable 2 (dummy)
> > interaction: interaction of variable 1 and variable 2
> >
> > tab var1 var2
> >
> > var1         |            var2
> >               |         0          1      |     Total
> > -----------+----------------------+----------
> >           0   |         0        122      |       122
> >           1   |       322        256      |       578
> > -----------+----------------------+----------
> >       Total   |       322        378      |       700
> >
> > Since no observations exist for var1==0&  var2==0 I can only include the
> > interaction and one of the variables (just to mention that: From a
> > theoretical sense it makes sense to do so): "probit var1 interaction"
> > Now I want to test if I can reject the hypothesis that
> > _b[var1]=_b[interaction]. If I use the standard command "test" it does not
> > account for the fact that these variables are related.
> >
> > Because of the nature of my variables I wanted to use the "contrast" command
> > but this only works if I'd use something like this before: "probit
> > var1##var2"  which is not solvable because of the above mentioned fact that
> > var1==0&  var2==0 does not exist in the data.
> >
> > Can anybody suggest another command that takes into account that the two
> > variables are interacted or has ideas on how to adjust the
> > "contrast"-command?
> >
> > Any help is greatly appreciated!
> >
> > Verena
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/