Re: failure of nlsList()

Douglas Bates (bates@stat.wisc.edu)
07 Jan 1998 09:00:27 -0600


Nathan Pace <nlpace@bigpace.med.utah.edu> writes:

> I'd appreciate help from anyone using nlsList().
>
> I called nlsList() using a parametrized data frame with starting values
> for the 4 parameters in the model equation. The parametrized data frame
> has no missing values. There are 1992 observations on 65 individuals
> (clusters). The model equation included the derivatives.
>
> The following error messages resulted:
>
> Error in nls(formula = formula, data = data, contr..: 47 NAs found in
> value
> Dumped
> Error was while calling subroutine "setup_nonlin"
> Error in nls(formula = formula, data = data, contr..: 44 NAs found in
> value
> Dumped
>
> .
> .
> .
>
> Error in nls(formula = formula, data = data, contr..: 21 NAs found in
> value
> Dumped
> Error was while calling subroutine "setup_nonlin"
>
> NAs were found in all observations for all individuals.
>
> Any thoughts about this problem?

The messages indicate that the NA's occurred in the evaluation of the
model function. Since they occurred in the routine "setup_nonlin"
which is the first time that the model function gets called when
fitting each cluster, I suspect a problem with the starting
estimates for the parameters.

You can try evaluating the model function explicitly on your data to
see the result. I think it will work to attach the parameterized data
frame then evaluate the model function or expression. You can contact
me if you want more details.

-- 
Douglas Bates                            bates@stat.wisc.edu
Statistics Department                    608/262-2598
University of Wisconsin - Madison        http://www.stat.wisc.edu/~bates/