[R] absurd computiation times of lme

Peter Dalgaard BSA p.dalgaard at biostat.ku.dk
Tue Oct 15 15:24:44 CEST 2002

Christof Meigen <christof at nicht-ich.de> writes:

> I think this is a bad approach and suggested to smooth the curves
> before. The approach of using (random = ~ age | Subject) or,
> as seen from looking at the log's, better (random = ~ age^0.15 | Subject),
> works as expected, but gives fits which are sometimes as far as
> 3 cm from the real measurements (while the measurement error
> is assumed to be about 0.5cm). These unusual decelerations are
> exactly what the wannabe-PhD is interested in.

This stuff can be tricky. We've had some success a couple of years back
with variants of models like

lme.obj <- lme(log(Height)~ns(sqrt(Age),knots=sqrt(c(0.25,.5,1,5)),
  Boundary.knots=c(0,sqrt(10))), random=~sqrt(Age)|ID,

This was also on children, albeit severely growth-retarded, and with a
less regular sampling than you seem to have. Notice that we felt it
was necessary to add some sort of correlation structure. One gotcha
which gave us convergence problems turned out to be due to the
starting value of corExp() being keyed to the *minimum* distance
between two measurements on the same individual. The really crucial
thing, however, was the need to work with transformations on both the
x and y axes in order to roughly stabilize the variance and avoid
"ringing" effects of fitting sharp gradients at small ages.
> Finally, the argument with the "too little data" does not apply
> to the set-up with with Berkeley Boys, with each one 31 measurements,
> where a 7-parameters spline basis random effect wouldn't converge
> within several hours.

Well, I'm sure that Doug and Jose would appreciate any insights that
could help them speed up the algorithms...

   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

More information about the R-help mailing list