R help - Sep 2001 - correlation predictors problem

R colleagues,

I want 
to get the correlation between the coefficients of a regression.
Everything seems OK, but when there are more than 3 regressors, the 
correlation matrix is completely wrong, as follows:
> x.glm <- glm( bascule  ~durP+  durI +durC+moyHzPB +moyHzIN,
x,family=binomial)
> summary(x.glm,correlation=T)
 
Call:
glm(formula = bascule ~ durP + durI + durC + moyHzPB + moyHzIN,
    family = binomial, data = x)
 .......
........
(Dispersion parameter for binomial family taken to be 1)
 
    Null deviance: 365.00  on 266  degrees of freedom
Residual deviance: 305.72  on 261  degrees of freedom
AIC: 317.72
 
Number of Fisher Scoring iterations: 3
 
Correlation of Coefficients:
        ( dP dI dC mHP
durP      1
durI         1
durC            1
moyHzPB .          1
moyHzIN ,
attr(,"legend")
[1] 0 ` ' 0.3 `.' 0.6 `,' 0.8 `+' 0.9 `*' 0.95 `B' 1 


I used binaries  rpm from  CRAN .
The same error is obtained with R1.2  on RH6.2 
and R.1.3.1-1 on RH7.0 (with the updated glibc2.2).
A similar  error is obtained with lm.

Thank you
-- 
Robert Espesser     
Laboratoire Parole et Langage  ESA 6057, CNRS
29 Av. Robert Schuman  13621 AIX    (FRANCE)
Tel: +33 (0)4 42 95 36 26   Fax: +33 (0)4 42 59 50 96
http://www.lpl.univ-aix.fr/~espesser
mailto:Robert.Espesser at lpl.univ-aix.fr
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On Tue, 11 Sep 2001, Robert Espesser -- CNRS wrote:
> R colleagues,
>
> I want
> to get the correlation between the coefficients of a regression.
> Everything seems OK, but when there are more than 3 regressors, the
> correlation matrix is completely wrong, as follows:
>
> > x.glm <- glm( bascule  ~durP+  durI +durC+moyHzPB +moyHzIN,
x,family=binomial)
> > summary(x.glm,correlation=T)
>
> Call:
> glm(formula = bascule ~ durP + durI + durC + moyHzPB + moyHzIN,
>     family = binomial, data = x)
>  .......
> ........
> (Dispersion parameter for binomial family taken to be 1)
>
>     Null deviance: 365.00  on 266  degrees of freedom
> Residual deviance: 305.72  on 261  degrees of freedom
> AIC: 317.72
>
> Number of Fisher Scoring iterations: 3
>
> Correlation of Coefficients:
>         ( dP dI dC mHP
> durP      1
> durI         1
> durC            1
> moyHzPB .          1
> moyHzIN ,
> attr(,"legend")
> [1] 0 ` ' 0.3 `.' 0.6 `,' 0.8 `+' 0.9 `*' 0.95 `B'
1
Notice the legend.  You don't have the correlation matrix but a
representation of it.

This is one of those things that certain people are keen on, and so
make the default for others.

You want

print(summary(x.glm,correlation=TRUE), symbolic.cor = FALSE)

I do think there should be a global option to turn this on, it being off
by default.  It's not even a documented argument in current R (but will be
in 1.4.0).

-- 
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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Robert Espesser -- CNRS <Robert.Espesser at lpl.univ-aix.fr> writes:
> R colleagues,
> 
> I want 
> to get the correlation between the coefficients of a regression.
> Everything seems OK, but when there are more than 3 regressors, the 
> correlation matrix is completely wrong, as follows:
....
> Correlation of Coefficients:
>         ( dP dI dC mHP
> durP      1
> durI         1
> durC            1
> moyHzPB .          1
> moyHzIN ,
> attr(,"legend")
> [1] 0 ` ' 0.3 `.' 0.6 `,' 0.8 `+' 0.9 `*' 0.95 `B'
1
Um, wrong in what way? 

If you just think this looks weird, blame Martin who wrote the code
for displaying correlation matrices like this. See the help for
symnum() for what it means. You can turn this off by passing
symbolic.cor=FALSE to print.summary.lm.

If the actual correlations are wrong, you need to tell us why you
think so. 

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._