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: omission of results due to multicollinearity

From   Conor Hughes <>
Subject   Re: st: omission of results due to multicollinearity
Date   Thu, 15 Jul 2010 11:13:45 +0700

Hi Sibel,
I don't know exactly the regression you're running, but hopefully I
can be specific enough to help out.  If Stata detects high enough
collinearity between two variables, it only drops one of the
variables, not both, though it decides which to drop on a more or less
arbitrary basis.  From a math standpoint, it doesn't really matter
which one is dropped, it mostly on affects your interpretation of the

For your case specifically, my guess would be that collinearity arises
for those ages specifically because 18-20 are the ages when people
typically go to college.  There's evidence that whether or not a
parent went to college is highly correlated with whether or not their
children go to college.  Parents' educational level doesn't matter so
much before the age of 18, since (in the US at least), you're required
to attend school until you're a legal adult, which is why you don't
see this collinearity at earlier ages.  You can check out the
correlations simply using the command -pwcorr-, for one.


On Wed, Jul 14, 2010 at 7:23 PM, Sibel S. <> wrote:
> Hi Statalist users,
> While running a regression to predict educational attainment for ages
> 7 through 20, some of the results for my age dummies and their
> interaction with parental schooling were omitted because of
> collinearity. I am trying to find out why this would happen only for
> ages 18 through 20. Does Stata drop both the variables when there is
> collinearity or one of them to improve the results? If it drops one of
> them, how can i find out with which other variable it has a high
> degree of multicollinearity. [I hope this is making sense]. Thanks.
> Sibel
> *
> *   For searches and help try:
> *
> *
> *

Conor Hughes
Mathematics and Economics
University of Chicago 2011

*   For searches and help try:

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