[R] different results with plot.lm vs. plot.lm(which=c(2))

[Greg Snow]

>From a quick look at the code it looks like when you ask for plot number 5 (included in default when 'which' is not specified), then the deviance residuals are replaced by the pearson residuals to be used in later computations.  So the difference that you are seeing is that one of the plots is based on deviance residuals and the other on pearson residuls.

It seems that there is a bug here in that, at a minimum, the label should be changed to indicate which residuals were actually used, or the code changed to continue to use the deviance residuals for plot 3 even when plot 5 is requested.

Does anyone else see something that I missed in how the residuals are replaced and used?

--
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111


> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of Effie Greathouse
> Sent: Wednesday, November 12, 2008 11:16 AM
> To: r-help at r-project.org
> Subject: Re: [R] different results with plot.lm vs. plot.lm(which=c(2))
>
> Hi Dr. Ripley--Sorry for the repost everybody.  The original message I
> sent
> never showed up in my inbox, so I thought it didn't get sent to the
> list.
>
> I'm running R 2.8.0, installed from a pre-compiled version, on Windows
> XP.
> When I type Sys.getlocale() at the R prompt, it returns:
>  "LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
> States.1252;LC_MONETARY=English_United
> States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252"
>
> Here's an example:
> bob <- seq(1:100)
> bob2 <- rgamma(100, 2, 1)*10+bob
> model<-glm(bob2 ~ bob, family=Gamma)
>
> Then enter:
> plot(model, which=c(3))
> to get the Scale-Location graph
>
> Then compare it to the Scale-Location graph when you run the following
> command and page through to the 3rd graph:
> plot(model)
>
> When I do this, I get different results -- some of the high values are
> different on each plot.  On my real data the difference is more severe
> than
> in this randomly generated example.  I'd be happy to supply my real
> data and
> R code if this smaller example isn't sufficient.  Thank you for any
> help!!
>
>
> On Wed, Nov 12, 2008 at 9:43 AM, Prof Brian Ripley
> <ripley at stats.ox.ac.uk>wrote:
>
> > Instead of re-posting the same message, please study the posting
> guide and
> > supply the information asked for, including a reproducible example.
> There is
> > no way we can help you unless you help us to help you.
> >
> >
> > On Wed, 12 Nov 2008, Effie Greathouse wrote:
> >
> >   I am running GLM models using the gamma family.  For example:
> >> model <-glm(y ~ x, family=Gamma(link="identity"))
> >>
> >> I am getting different results for the normal Q-Q plot and the
> >> Scale-Location plot if I run the diagnostic plots without specifying
> the
> >> plot vs. if I specify the plot ... e.g., "plot(model)" gives me a
> >> different
> >> Normal Q-Q graph than "plot(model, which=c(2))".  The former gives
> data
> >> points distributed in a quadratic pattern, while the latter gives
> data
> >> points more or less along the 1:1 line.  Shouldn't these two
> commands be
> >> giving me the same exact graphs?  I have read the documentation on
> plot.lm
> >> and searched the help archives, but I am still learning GLM's and
> I'm not
> >> very familiar with understanding diagnostic plots for GLM's, so any
> help
> >> would be much appreciated!
> >>
> >>        [[alternative HTML version deleted]]
> >>
> >> ______________________________________________
> >> R-help at 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<http://www.r-
> project.org/posting-guide.html>
> >> and provide commented, minimal, self-contained, reproducible code.
> >>
> >
> > --
> > Brian D. Ripley,                  ripley at 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
> >
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at 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.