Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

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

Re: st: interactions of treatment in treatreg

From   Austin Nichols <>
Subject   Re: st: interactions of treatment in treatreg
Date   Tue, 6 Jul 2010 08:33:22 -0400

Thanos Mergoupis:
-treatreg- is for a single endogenous binary explanatory variable
(treatment) and a continuous outcome, and does not allow interactions
with treatment.  If you had used factor variables, you would have
treatment appearing in two places and it would not converge
(probably).  If you hard coded an interaction and included it in the
main equation, you have tricked it into giving you biased and
inconsistent estimates.  If you want an interaction, use -ivregress-
or -ivreg2- (SSC) and recognize that now you have at least 2
endogenous explanatory variables (treatment and its interaction) so
you need more excluded instruments.

On Tue, Jul 6, 2010 at 6:32 AM, Maarten buis <> wrote:
> --- On Tue, 6/7/10, Thanos Mergoupis wrote:
>> Has anybody noticed that in treatreg, when using
>> interactions of the treatment with variables in the
>> outcome equation, the postestimation predict
>> command that conditions on treatment (or non-treatment),
>> does not condition the interaction terms on treatment (or
>> non-treatment)? It considers the interaction terms just
>> like any other variable, ignoring the fact that they
>> are interactions with the treatment. Does anybody know of
>> any way to make stata recognize them as interactions with
>> the treatment so as to calculate the correct predicted values?
> In general yes, see: -help fvvarlist-. However, I don't know
> if this makes sense in the -treatreg- case.
> Hope this helps,
> Maarten
*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index