Hi,

I turn to the help list after spending hours on the package help, looking
online, asking colleagues and trying various versions on R itself.

The work I need help on involves a confidentiality contract, so I am limited
in details on the specific model I used. However, my problem is really
simple in essence, and is described below.

The theoretical model for response is roughly :
y = (intercept+linear effects of certain covariates) + alpha * exp (beta*t).
(with beta generally believed to be positive, i.e., y begins at some
horizontal asymptote, and increases unboundedly as time passes)

There is one grouping variable (let's call it "ID").

Most of the fixed effects need to have a random component, i.e. for
different ID's they would vary around the fixed mean.

Here are my questions:

1. Do nlme/nlmer accept on the RHS of the main formula, only SelfStart-type
objects? (I assume yes, because I tried to code the RHS directly in the
formula with no success).

2. If so, is there any SelfStart object I can take "off the shelf" and use
for this - or must I create my own? (would need guidance to do the latter)

3. In any case, can you instruct me about the proper syntax for the main,
fixed and random formulae - or for the seemingly-niftier nlmer interface?

Here are some of the errors I get:
Error in nlmeCall[[i]] <- NULL : subscript out of bounds

Error in nlme.formula():
  element 1 is empty;
   the part of the args list of 'is.numeric' being evaluated was:
   (start)

By the way, the data integrity is fine. I have no problem running lmer on
the same dataset. I use R 2.8.1 for Windows and the latest package versions.

Thanks in advance, Assaf

        [[alternative HTML version deleted]]

______________________________________________
R-help@r-project.org 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.

Reply via email to