Re: [Rd] a bug in glm.fit() (PR#7947)

From: Prof Brian Ripley <ripley_at_stats.ox.ac.uk>
Date: Thu 16 Jun 2005 - 06:06:49 GMT

What is the bug?

This is the same model: the `intercept' term affects the null model, not the actual model. Just look at all the output.

On Thu, 16 Jun 2005 jyzz88@gmail.com wrote:

> glm.fit() gave me the same AIC's regardless of TRUE or FALSE intercept option.
>
>> myX <- as.matrix(1:10)
>> myY <- 3+5*myX
>> foo <- glm.fit(x=myX, y=myY, family = gaussian(link = "identity"), intercept=TRUE)
>> foo$aic
> [1] 38.94657
>> foo <- glm.fit(x=myX, y=myY, family = gaussian(link = "identity"), intercept=FALSE)
>> foo$aic
> [1] 38.94657
>> AIC(lm(myY~0+myX, data=data.frame(myY,myX)))
> [1] 38.94657
>> AIC(lm(myY~1+myX, data=data.frame(myY,myX)))
> [1] -650.9808

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
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-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Received on Thu Jun 16 16:09:20 2005

This archive was generated by hypermail 2.1.8 : Mon 20 Feb 2006 - 03:21:09 GMT