Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Re: your stata query

From   "Elizabeth Whiting" <[email protected]>
To   [email protected], [email protected]
Subject   st: Re: your stata query
Date   Thu, 16 Jun 2005 08:46:54 +0100

Hi Sarah

Thanks for this information. I think that is exactly the problem, because I
am trying to enter the same variables in the model more than once (like in
your first example). Is the model correct then? Can you interpret it in the
same way, even though some of the variables appear to be collinear? The
other problem is that when I enter the same model into gllamm it won't run
at all because it says the variables are all collinear. It seems strange
why it will run in STATA but not in gllamm.

Thanks for your help


                      [email protected]                                                                                                
                           To:       Elizabeth Whiting/SRD/SESAG/LONDON/ONS@ONS                                    
                      13/06/2005 21:46          Subject:  your stata query                                                              

Did anyone ever get back to you on this?  I don't want to post a
response to the list, because I haven't been paying attention to see if
anyone already responded!  Just in case no one did...

This typically happens for me when I have multiple interaction terms
with some of the same variables in them OR if I use -xi- and also
include the main effects in the model.  In the former case, if I am
doing interactions, say, by time and have something like,

xi: xtreg depress i.time*SES i.time*socsupp , i(childid)

then I will get a message about the time variables being dropped due to
multicollinearity - because they are included twice in the model.
That's why the model still seems to run fine and all the variables seem
to be included - because they are!

In the second case, if you use -xi- and forget that it includes the main
effects automatically and type something like:

xi: xtreg depress  i.time SES socsupp i.time*SES i.time*socsupp ,

then the time dummies, SES and socsupp would all get dropped due to
multicolliearity, just because they are in there too many times.

Many people recommend avoiding -xi- for that reason.

Hope this helps.



I am running a logistic model with lots of 2 and 3 way interactions in it.
In the output STATA produces a list all of the variables which it says have
been dropped due to multicollinearity (which is basically all of them), but
it then goes on to produce the model with parameter estimates for all of
the variables that it says it has dropped. Does anyone know what this
means? Can I interpret the model the same as you normally would?

In addition when you run the same thing in gllamm with a second level
specified it again lists all the variables, which have been dropped due to
multi-collinearity, but in this case it says it can't run the model and it
then stops without producing any estimates. Does anyone know how you get
around this problem? Why will it run in logit regression but not in gllamm?

Thanks for your help


Sarah A. Mustillo, Ph.D
Department of Psychiatry and Behavioral Sciences
Duke University School of Medicine
Box 3454
Durham NC 27710

919 687-4686 x231


On entering the GSi, this email was scanned for viruses by the Government
Secure Intranet (GSi) virus scanning service supplied exclusively by
Energis in partnership with MessageLabs.

Please see
for further details.

In case of problems, please call your organisational IT helpdesk

For the latest data on the economy and society 
consult National Statistics at

Please Note:  Incoming and outgoing email messages
are routinely monitored for compliance with our policy
on the use of electronic communications
Legal Disclaimer  :  Any views expressed by
the sender of this message are not necessarily
those of the Office for National Statistics

The original of this email was scanned for viruses by the Government Secure Intranet (GSi) virus scanning service supplied exclusively by Energis in partnership with MessageLabs.

On leaving the GSi this email was certified virus-free
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index