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: st: Explaining apparent inconsistency in regression coefficients

From   Nick Cox <>
Subject   Re: st: Explaining apparent inconsistency in regression coefficients
Date   Fri, 8 Apr 2011 20:45:41 +0100

Your relation Y3 = Y1 + Y2 does not carry over to your models. Lots of
zeros on one Y could lever down the corresponding regressions.

You should be able to plot everything on a single graph to see what is
going on.


On Fri, Apr 8, 2011 at 8:23 PM, Hussain Samad <> wrote:

> This is not exactly a STATA question except for the program is run in STATA. I
> am running a say OLS (not TOBIT) for three outputs:
> log(1+Y1)= alpha + beta*X + epsilon
> log(1+Y2)= alpha + beta*X + epsilon
> log(1+Y3)= alpha + beta*X + epsilon
> where Y1= farm income, Y2=nonfarm income, and Y3=total income=Y1+Y2, and X is a
> dummy intervention variable. Y1 or Y2 can be 0 but Y3 cannot. And these are the
> beta values I'm getting: 0.45 for Y1, 0.35 for Y2 and 0.21 for Y3. My
> question: why is the beta for Y3 is smaller than the other two? Shouldn't it be
> in between the other two betas? Any explanation will be highly appreciated.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index