[R] [SPAM] - RE: REML with random slopes and random intercepts giving strange results - Bayesian Filter detected spam

Peter Dalgaard p.dalgaard at biostat.ku.dk
Wed Aug 16 14:45:40 CEST 2006


"Simon Pickett" <S.Pickett at exeter.ac.uk> writes:

> sure, thanks again.

What is the order of magnitude of newwt? If it is large and with a
small variance, then the -1 correlation might be an artifact of
extrapolating the individual lines to zero. Try centering newwt by
subtracting its mean.

> > summary(m2)
> Linear mixed-effects model fit by REML
> Formula: change.wt ~ newwt + (newwt | id)
>    Data: grow
>        AIC       BIC   logLik MLdeviance REMLdeviance
>  -6203.178 -6164.462 3107.589  -6239.374    -6215.178
> Random effects:
>  Groups   Name        Variance   Std.Dev.  Corr
>  id       (Intercept) 1.0868e-02 0.1042482
>           newwt       4.7069e-05 0.0068606 -1.000
>  Residual             1.4236e-02 0.1193136
> # of obs: 4688, groups: id, 485
> 
> Fixed effects:
>                Estimate  Std. Error   DF t value  Pr(>|t|)
> (Intercept)  5.5692e-01  6.4189e-03 4686  86.761 < 2.2e-16 ***
> newwt       -3.2382e-02  4.5962e-04 4686 -70.455 < 2.2e-16 ***
> ---
> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> 
> Correlation of Fixed Effects:
>       (Intr)
> newwt -0.954
> 
> > Can you provide the summary(m2) results?
> >
> >> -----Original Message-----
> >> From: Simon Pickett [mailto:S.Pickett at exeter.ac.uk]
> >> Sent: Wednesday, August 16, 2006 7:14 AM
> >> To: Doran, Harold
> >> Cc: r-help at stat.math.ethz.ch
> >> Subject: [SPAM] - RE: [R] REML with random slopes and random
> >> intercepts giving strange results - Bayesian Filter detected spam
> >>
> >> Hi again,
> >> Even stranger is the fact that the coefficeints (the slope)
> >> and the intercepts are not independent, in fact they are
> >> directly inversely proportional (r squared = 1).
> >> This means that that there isnt a random slope and intercept
> >> for each individual (which is what I wanted), but straight
> >> line that pivots in the middle and will change from
> >> individual to individual. Is there a problem with the way I
> >> have structured the random model or a deeper problem with lmer()?
> >> here is the code I used
> >> m2 <- lmer(changewt ~ newwt+(newwt|id), data = grow)
> >> coef(m2)
> >> Any suggestions very much appreciated,
> >> Simon
> >>
> >>
> >> > I don't this is because you are using REML. The BLUPs from a mixed
> >> > model experience some shrinkage whereas the OLS estimates would not.
> >> >
> >> >> -----Original Message-----
> >> >> From: r-help-bounces at stat.math.ethz.ch
> >> >> [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of
> >> Simon Pickett
> >> >> Sent: Tuesday, August 15, 2006 11:34 AM
> >> >> To: r-help at stat.math.ethz.ch
> >> >> Subject: [R] REML with random slopes and random intercepts giving
> >> >> strange results
> >> >>
> >> >> Hi everyone,
> >> >> I have been using REML to derive intercepts and
> >> coeficients for each
> >> >> individual in a growth study. So the code is
> >> >> m2 <- lmer(change.wt ~ newwt+(newwt|id), data = grow)
> >> >>
> >> >> Calling coef(model.lmer) gives a matrix with this
> >> information which
> >> >> is what I want. However, as a test I looked at each
> >> individual on its
> >> >> own and used a simple linear regression to obtain the same
> >> >> information, then I compared the results. It looks like the REML
> >> >> method doesnt seem to approximate the two parameters as
> >> well as using
> >> >> the simple linear regression on each individual
> >> separately, as judged
> >> >> by looking at graphs.
> >> >> Indeed, why do the results differ at all?
> >> >> Excuse my naivety if this is a silly question.
> >> >> Thanks to everyone for replying to my previous questions,
> >> very much
> >> >> appreciated.
> >> >> Simon Pickett
> >> >> PhD student
> >> >> Centre For Ecology and Conservation
> >> >> Tremough Campus
> >> >> University of Exeter in Cornwall
> >> >> TR109EZ
> >> >> Tel 01326371852
> >> >>
> >> >> ______________________________________________
> >> >> R-help at stat.math.ethz.ch 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.
> >> >>
> >> >
> >>
> >>
> >> Simon Pickett
> >> PhD student
> >> Centre For Ecology and Conservation
> >> Tremough Campus
> >> University of Exeter in Cornwall
> >> TR109EZ
> >> Tel 01326371852
> >>
> >>
> >
> 
> 
> Simon Pickett
> PhD student
> Centre For Ecology and Conservation
> Tremough Campus
> University of Exeter in Cornwall
> TR109EZ
> Tel 01326371852
> 
> ______________________________________________
> R-help at stat.math.ethz.ch 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.
> 

-- 
   O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907



More information about the R-help mailing list