Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: x predicts failure perfectly?


From   "Petersen, Karen " <k.petersen@vanderbilt.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: x predicts failure perfectly?
Date   Mon, 5 Apr 2004 13:36:49 -0500

I have run into a perplexing problem.  I am using logit to analyze an outcome 
variable (mid)

logit mid illibdem1 limdyn1 absdyn1 auth1 pers1 cwpceyrs_werner rel_cap 
econ_dep1 econ_dep2 rel_energy allies if cat2_2==11 & exclude2==0 & 
initiate~=2, cluster(dyad_id)  

and get the following message about one of my explanatory variables (absdyn1):

note: absdyn1 != 0 predicts failure perfectly
      absdyn dropped and 68 obs not used


Given that message, I ran a cross tab expecting to find a perfect 
correlation.  Instead, I found:
 
. tab x y if cat2_2==11 & exclude2==0 & initiate~=2

           |           y
         x |         0          1 |     Total
-----------+----------------------+----------
         0 |    41,539        326 |    41,865 
         1 |     1,829         16 |     1,845 
-----------+----------------------+----------
     Total |    43,368        342 |    43,710 


I would just force Stata to keep the variable using "asis", but I am also 
running the Clarify ado commands-which will not work with "asis."  I could 
change my reference category to "absdyn" rather than the category I currently 
exclude "libdem1."  (The first 5 explanatory variables are part of a set of 6 
mutually exclusive categorical indicators.) However, I am not sure if that is 
the best solution.

Does anyone know why I get the above error message when the variable does not 
appear to actually be a "perfect predictor"?

Thanks for your help,

Karen K. Petersen 
Vanderbilt University 
Department of Political Science 
Box 1817 Station B 
Nashville, Tennessee 37235 
US 
615-322-6226
k.petersen@vanderbilt.edu
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index