R help - Jan 2003 - as.formula(string) and augPred in lme

Using formulas constructed from strings only
partially works for me in lme:

library(nlme)
data(Orthodont)

fm2<-lme(as.formula("distance~age"),data=Orthodont,random=~1|Subject)
summary(fm2) # works
augPred(fm2) # fails
#Error in inherits(object, "formula") : 
#Argument "object" is missing, with no default

I assume that my use of as.formula is wrong, but what's the right way?
Or at least a workaround?


Dieter Menne

On Thu, 30 Jan 2003, Dieter Menne wrote:
> Using formulas constructed from strings only
> partially works for me in lme:
> 
> library(nlme)
> data(Orthodont)
> 
>
fm2<-lme(as.formula("distance~age"),data=Orthodont,random=~1|Subject)
> summary(fm2) # works
> augPred(fm2) # fails
> #Error in inherits(object, "formula") : 
> #Argument "object" is missing, with no default
> 
> I assume that my use of as.formula is wrong, but what's the right way?
> Or at least a workaround?
The error is in predict.lme, I think.  That has

eval(as.call(mCall[["fixed"]][-2]))

and I think

eval(mCall[["fixed"]])[-2]

may be the intention.

Workaround:  use lme(distance ~ age, data=Orthodont,random=~1|Subject)
or if you need to construct a formula, use

form <- as.formula("distance ~ age")
fm2 <- eval(substitute(lme(form, data=Orthodont, random=~1|Subject),
list(form=form)))

or alter predict.lme as indicated).

-- 
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