On Tue, 13 May 2008, Diego Cesar wrote:
Hello guys, i looked over the archive files and found nothing about this
kind of error.
It's a feature.
I have a database of 33 elements described in 8 variables, i'm using the
Leave-One-Out iterative process
to take one of the elements to be the test element and make a regression
with the other 32 and then
I try to predict the clas of the element out.
I'm using this call as a part of a Leave-One-Out experiment:
glmresult <- glm(Y ~ x1+x2+x3+x4+x5+x6+x7+x8, family=binomial,
control=glm.control(epsilon = 1e-8, maxit = 100, trace = FALSE))
but when i look at the coefficients with:
coef(glmresult)
Everything goes well (i get all the coefficients) until the 27th element
which gets a NA as a coefficient to variable x8
The vectors Y, x1,x2,x3,x4,x5,x6,x7,x8 have the same length and there isn't
any missing data.
don't know what to do right now as i already reviewed the code a lot and i'm
starting to think the problem is in the
data.
Correct, collinearity aka extrinsic aliasing. x8 is not linearly
independent of (1, x1 ... x7) on your dataset and so is dropped from the
model.
Diego Cesar
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
--
Brian D. Ripley, [EMAIL PROTECTED]
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
