[R-lang] p-values for factors using lmer & mcmc
T. Florian Jaeger
tiflo at csli.stanford.edu
Thu Sep 6 06:22:37 PDT 2007
On 9/6/07, Andy Fugard <a.fugard at ed.ac.uk> wrote:
>
> T. Florian Jaeger wrote:
> >
> > On 9/3/07, *Andy Fugard* <a.fugard at ed.ac.uk <mailto:a.fugard at ed.ac.uk>>
> > wrote:
> >
> > [...] Faraway (2006, p. 13) mentions in passing
> > that "We would normally avoid using the t-tests for the levels of
> > qualitative predictors with more than two levels." His example is
> for
> > Gaussian multiple regression; perhaps that's important?
> >
> > that is because the levels of a qualitative predictor are often
> > collinear. collinearity is bad for tests that use the standard error
> > estimate of the coefficient. so, those issues should go away after you
> > center your predictors and go through other measures to remove/reduce
> > collinearity in your model.
>
> Aha! Is there a standard reference for this?
Agresti 2002 Categorical Data Analysis discusses Wald statistics, and I
think he also discusses collinearity. But let me be clear that I haven't
read that the reason for unreliability of Wald Statistics is exclusively
collinearity. I just meant to say that since the Wald Statistics is based on
the standard error of the coefficient it is going to be vulnerable to
collinearity.
Florian
Many thanks,
>
> Andy
>
> --
> Andy Fugard, Postgraduate Research Student
> Psychology (Room F15), The University of Edinburgh,
> 7 George Square, Edinburgh EH8 9JZ, UK
> Mobile: +44 (0)78 123 87190 http://www.possibly.me.uk
> _______________________________________________
> R-lang mailing list
> R-lang at ling.ucsd.edu
> https://ling.ucsd.edu/mailman/listinfo.cgi/r-lang
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://ling.ucsd.edu/pipermail/r-lang/attachments/20070906/052d7d4a/attachment.htm
More information about the R-lang
mailing list