Re: st: Order of variables in stcox model

[Steven Samuels <sjsamuels@gmail.com>]

I"ll just point out that if the poster had followed the directions in the FAQ (section 3.3) and showed us exactly what he had typed what Stata had responded, this correspondence would have been a lot shorter.


Steve

On Aug 15, 2011, at 11:06 AM, Maarten Buis wrote:

On Mon, Aug 15, 2011 at 4:51 PM, Brigham Whitman  wrote:
> Hmm, OK. Thank you for the advice.  The comment about
> multicolinearity: I have already checked the Pearson's correlation
> with "pwcorr" and none of the variables have a correlation above .7.
> When I run the model, however, a lot of the variables are dropped for
> whatever reason, and I just get a period (".") instead.

That is the reason why the order of variables mattered in your model.
Remember to multicolinearity is not only about bivariate correlations,
no variable should be an exact linear combination of any of the
remaining variables. Classic examples are a set of dummies
representing a categorical variables, where you must exclude one
reference category, or age period and cohort, the cohort (year of
birth) is the period (year in which survey was held) - age, so you
cannot simultaneously estimate the effect of all three (at least not
without some special models that include some heavy assumptions).

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/