[nlme] BLUPs for a new subject in a fitted lme model?

I am seeking for a method to calculate, given a fitted lme model
and some data for a subject, the random effects predictors
for this subject. I can only find predictors for the subjects used in
creating the fit. Of course I could just add the subject and redo the fit.
But I want to avoid just this refitting. 

Thanks for help
wbk

On Fri, 16 Aug 2002, Wilhelm B. Kloke wrote:

I am seeking for a method to calculate, given a fitted lme model
and some data for a subject, the random effects predictors
for this subject. I can only find predictors for the subjects used in
creating the fit. Of course I could just add the subject and redo the fit.
But I want to avoid just this refitting.

Thanks for help
wbk

Once you've a lme model, you don't need to stimate new random effects for
new subjects.

As stated by Pinheiro & Bates:
"Fixed affects are parameters associated with an entire population, or
with repeatable levels of experimental factors.
Random effects are instead associated with experimental units drawn at
random from a population."

Maybe you'd like to see Pinheiro & Bates' book for further details.

That's not the whole story.  What (as I understand it) Wilhelm has are not
just new subjects, but new observations on new subjects.  Then there are
BLUPs associated with those new subjects.  Think of an enlarged data set
with the old and the new subjects: you can apply the lme at a given set of
parameter values and find the BLUPs for all the subjects.

It is equally true that new observations on the old subjects will change
the BLUPs.

I don't know how to do this with lme without refitting, although the
pieces needed must be in the package.


On Fri, 16 Aug 2002, Ulises Mora Alvarez wrote:

On Fri, 16 Aug 2002, Wilhelm B. Kloke wrote:

I am seeking for a method to calculate, given a fitted lme model
and some data for a subject, the random effects predictors
for this subject. I can only find predictors for the subjects used in
creating the fit. Of course I could just add the subject and redo the fit.
But I want to avoid just this refitting.

Thanks for help
wbk

Once you've a lme model, you don't need to stimate new random effects for
new subjects.

As stated by Pinheiro & Bates:
"Fixed affects are parameters associated with an entire population, or
with repeatable levels of experimental factors.
Random effects are instead associated with experimental units drawn at
random from a population."

Maybe you'd like to see Pinheiro & Bates' book for further details.

That's not the whole story.  What (as I understand it) Wilhelm has are not
just new subjects, but new observations on new subjects.  Then there are
BLUPs associated with those new subjects.  Think of an enlarged data set
with the old and the new subjects: you can apply the lme at a given set of
parameter values and find the BLUPs for all the subjects.

It is equally true that new observations on the old subjects will change
the BLUPs.

I don't know how to do this with lme without refitting, although the
pieces needed must be in the package.


On Fri, 16 Aug 2002, Ulises Mora Alvarez wrote:

On Fri, 16 Aug 2002, Wilhelm B. Kloke wrote:

I am seeking for a method to calculate, given a fitted lme model
and some data for a subject, the random effects predictors
for this subject. I can only find predictors for the subjects used in
creating the fit. Of course I could just add the subject and redo the fit.
But I want to avoid just this refitting.

Thanks for help
wbk

Once you've a lme model, you don't need to stimate new random effects for
new subjects.

As stated by Pinheiro & Bates:
"Fixed affects are parameters associated with an entire population, or
with repeatable levels of experimental factors.
Random effects are instead associated with experimental units drawn at
random from a population."

Maybe you'd like to see Pinheiro & Bates' book for further details.

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

In the simple model Y = mu + U + e, where U and e have variances Vu and
Ve, the BLUP for a subject with observation Y is

E(U|Y) = Vu  * (Y - mu)
        ----
       (Vu+Ve)

Similarly for more complicated models, i.e. the BLUP is a simple function
of the variance components. I'm not sure whether this answers the original
question, but it must be relevant and nobody has mentioned it so far.