--- Michael McCulloch <[email protected]> wrote:
> I'm comparing forward vs. backward options for swaic, after stcox.
> Different models are achieved. Is there a rule-of-thumb about which
> to use?
Don't add two variables that measure the same thing, that way you are
controlling a variable for itself.
Do control for a variable that causes both the explanatory variable of
interest and the dependent variable. Classic example is the quality of
the summer in the previous year in a rural area causes both the number
of storks to increase (more food, more baby storks survive) and causes
the number of humans babies to increase (good harvest, more positive
outlook on the future, and don't forget the nice harvest party). If you
don't control for the weather during previous summer you might think
that the number of storks determine the number of babies.
If two explanatory variables are just correlated, then it depends
whether you want the total effect, i.e. the direct effect plus indirect
effect, or the direct effect. The total effect usually makes more
descriptive sense to me, but the direct effect is closer to a causal
story.
Which scenario fits which variable in your model is determined by your
substantive knowledge of the problem you are investigating.
Hope this helps,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
___________________________________________________________
All new Yahoo! Mail "The new Interface is stunning in its simplicity and ease of use." - PC Magazine
http://uk.docs.yahoo.com/nowyoucan.html
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
