Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: huge test statistic with -logit-


From   "Verkuilen, Jay" <JVerkuilen@gc.cuny.edu>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   st: RE: huge test statistic with -logit-
Date   Mon, 7 Dec 2009 11:57:08 -0500

Sue wrote:

<<<I'm seeing some huge test statistics that I can't seem to explain. <snip> Does anyone have an idea why
this might be happening?>>>

You almost certainly have multicollinearity or near-perfect prediction, especially with those interactions. Interactions of continuous or a continuous and categorical variable are notorious for being collinear with their main effects and interactions of categorical variables frequently create perfect prediction. In a true perfect prediction case (when knowing a value of the IV perfectly predicts the DV), Stata will kick a variable out. However, if you don't *quite* have perfect prediction the estimates and standard errors will become unstable in exactly the way you describe. 

Nick Cox already indicated that you need to take some serious steps back to a much simpler model. In short: Clean out that kitchen sink! 

Jay

*
*   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/



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