[R] two questions on nlme: error messages and nested variance

J josh8912 at yahoo.com
Fri Oct 1 09:21:31 CEST 2004


Hello:
Im hoping the list can shed light on two items.  I am
using an nlme model to fit a logistic function and
have these questions:

1)  The model works fine using varIdent(~1|factor) and
also using
varComb(varIdent(~1|factor),varFixed(~covariate)), but
not when varFixed(~covariate) is used alone.  Using
VarFixed alone gives the message: 

Error in recalc.varFunc(object[[i]], conLin) : dim<- :
dims [product 153] do not match the length of object
[846]  In addition: Warning message: longer object
length is not a multiple of shorter object length in:
conLin$Xy * varWeights(object).

Can anyone shed light on what this message means and
how I might go about fixing the problem?  I am still
new to all this, so please explain with that in mind.

2)  In my data, the mean of the variance of the
within-group errors is dependent on a covariate, and
in addition the variance of the variance is also
dependent on a covariate.  To model this would one use
something like: 
varIdent(~varIdent(~1|covariate)|covariate)?  I get
the same error message as above when I try this or if
I try: varIdent(~varFixed(~covariate)|factor).  Is
there a better way to do such nesting?  Is it even
possible?  And again, what does the error mean?

Thanks much in advance.  John




More information about the R-help mailing list