[R-sig-ME] partial residuals or another method to check nonlinear model spec.? (emails keep bouncing)

[Seth Myers]

Hi,

Hi,

This might be a double post because my previous entry bounced.  If so,
I apoogize.

I have not found a method to extract partial residuals plots given a
nonlinear model employed in gnls.

My dependent variable is bound between 0 and 1 but can take on
fractional values such as 0.75, and a logistic function of a few
variables seems to do well when predictions are plotted against the
dependent variable.

Right now, the terms are simply entered as being linear within the
model (I might not be stating that precisely, but the model is simply

y=1/(1+exp1(b0 +b1*X1))

if only one independent variable were included).

I suppose I could use splines and see what transformations or other
models are suggested by them.  Given that the dependent is not binary,
I might be able to transform the dependent variable using a link
function and then simply run a linear model and extract partial
residuals from it.  But I am not sure if this can be justified. Is
there a better and/or easier approach someone could recommend?