[R] changes in glm?

Thomas Lumley thomas at biostat.washington.edu
Thu Nov 30 22:12:54 CET 2000

On Thu, 30 Nov 2000, G.Janacek wrote:

> I am running some simulations which involve fitting a glm with 
> Binomial error ( in fact Bernoulli) and logistic link. Running R 0.63 
> I have no problems. However I need more speed so migrated to our UNIX 
> boxes which run
> Version 0.90.1  (December 15, 1999)
> Here I get an error
> Error in (if (is.empty.model(mt)) glm.fit.null else glm.fit)(x = X, y 
> = Y,  : y must be univariate unless binomial
> which has baffled me.
> So I as your indulgence and ask
> (1) What if anything has changed in glm from 0.63 to 0.90?

I shouldn't think anyone can remember -- that's a long time ago.

> (2) Is glm sensative to changes of word length?

Shouldn't be.  I have used in on both 32-bit and 64-bit Solaris.

> (3) Given making R on an Alpha is beyond me (so is a problem) should 
> I migrate to a later version of R or does this bring more problems in 
> its wake?

Yes, definitely update. Many many bugs were fixed before 1.0.0.  You won't
get any support for pre1.0 versions.  The newest version is 1.1.1, and
while a shiny new 1.2 is due out in a couple of weeks we expect quite a
lot of people to stick with 1.1.1 for the rest of the Northern hemisphere
academic year.

> (4) Does anyone recognise the error and if so I would appreciate some 
> info. I'm sure it is my stupidity but ..

It's hard to tell without knowing exactly what you did to get the error.
The error means that glm() thinks you gave it a matrix rather than a
vector as the response (outcome, dependent) variable.

Remember, the purpose of a bug report is for us to reproduce the bug. This
isn't being difficult -- we usually can't fix bugs unless we can reproduce

Bug reports get fixed a lot faster if they are of the form

	"I did X. Y happened. I think Z should have happened"


Thomas Lumley
Assistant Professor, Biostatistics
University of Washington, Seattle

r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list