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: Re: st: Interaction terms

From   lreine ycenna <>
Subject   Re: Re: st: Interaction terms
Date   Wed, 4 May 2011 11:00:19 +0100

Thank you.

 I tried the factorial method, however, I have very large
coefficients,high SE, especially when the group dummy G1 is interacted
with all the variables and on sub-sample (e.g. if year < 2000 and >
1995).so my results are mostly insignificant.

(1) I wonder if it's because the second half of my variables are the
bi-products of first half, even though I'm meant to treate these
bi-product variables as individual variables. As I gradually add more
variables, I also have more bi-product variables.

e.g. (a) regress y ov edu wealth gd eduxwealth eduxgd weathxgd ovxedu
ovxwealth ovxgd. ovxeduxwealth ovxeduxgd ovxwealthxgd.

In this case, would it biase my result to include so many bi-product
variables? If so, does it make sense to run all the bi-products
separately on a single regression? e.g. (b) regress y ov eduxwealth
eduxgd weathxgd ovxeduxwealth ovxeduxgd ovxwealthxgd. And then compare
the coefficients in (b) with (a).

(2) I notice that "Regress y ov edu wealth gd ovxedu ovxwealth ovxgd
if G1==1" produces different/ smaller coefficients and SE from that of
the ==G1 command. Would it be incorrect to use the if G1==1 method
instead of regress y i.G1##c.(ov edu wealth gd ovxedu ovxwealth
ovxgd)? I don't quite understand the difference.

(3) what command do I need to divide my data into 5 year panels?

year  Country
1997 UK
1998 UK
1999 UK
2000 UK
2001 UK
2002 UK
2003 UK
2004 UK
2005 UK
2006 UK
2007 UK
2008 UK
1997 AU
1998 AU
1999 AU
2000 AU
2001 AU
2002 AU
2003 AU
2004 AU
2005 AU
2006 AU
2007 AU
2008 AU


On 3 May 2011 18:11, Christopher Baum <> wrote:
> <>
> I have two dummies for country groups G1 and G2.
> In the stata command, I could (1) interact the term so it becomes:
> regress y ov edu wealth eduxG2 wealthxG2 ovxeduxG2 ovxwealthxG2
> or (2) I could run: regress y ov edu wealth ovxedu ovxwealth if G1 ==1.
> Would the 2nd method be wrong? What's the difference between (1) and (2)?
> Obviously not the same model. If you wanted to fully interact the G1/G2 split with the model
> regress y ov edu wealth ovxedu ovxwealth
> it would be
> regress y i.G2##(ov edu wealth ovxedu ovxwealth)       (1)
> If you then did
> regress y ov edu wealth ovxedu ovxwealth if G1            (2)
> you would get the same point estimates in (1) for the G1 countries as you do in (2), but the standard errors would differ.
> In the pooled regression (1) you are constraining the sigma^2 to be equal across G1 and G2, whereas in (2) and the equivalent
> model you could run for group G2 that constraint is not applied.
> Note that the fully interacted model needs G2 by itself as well as interacted. Your model above with the "xG2" variables does not do that.
> Kit
> Kit Baum   |   Boston College Economics & DIW Berlin   |
>                              An Introduction to Stata Programming  |
>   An Introduction to Modern Econometrics Using Stata  |
> *
> *   For searches and help try:
> *
> *
> *

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index