[R] How to solve an NLME problem?

[Ben Bolker]

David C Blouin <dblouin <at> lsu.edu> writes:

>  I have a nonlinear model where I want to include random
> coefficients for a sample of random SUBJECTS. The fixed effects part
> of the model is Y ~ B + (T - B) / (1 + 10**(X - C)) and I would like
> to include random coefficients b for B, t for T, and c for C. The
> model is a fairly well-known three-parameter log-inhibitor model
> where Y is a measure of growth at X = log(CONCENTRATION), B is the
> unknown lower asymptote, T is the unknown upper asymptote, and C is
> the unknown log(IC50), where IC50 is the half maximal inhibitory
> concentration. I do not know how to solve this problem in R, where I
> am assuming NLME contains the appropriate methodology.
> [[alternative HTML version deleted]]

  something like

  nlme(Y ~ B + (T - B) / (1 + 10**(X - C)),
       random=B+T+C~group,
       data=your_data,
       start=list(fixed=c(B=something,T=something2,C=something3)))

 Follow-ups to r-sig-mixed-models at r-project.org, please ...