I'm trying to use nonlinear regression to estimate model parameters for a constant rate birth-death process conditioned on the fact that the population has not gone extinct. Because the birth and death parameters obtained from standard application of gnls() appear to be biased, I'd like to write a new varFunc class for gnls() that accommodates the expected increase in variance with time. The help documentation for varClasses notes that "Users may define their own varFunc classes by specifying a constructor function and, at a minimum, methods for the functions coef, coef<-, and initialize. For examples of these functions, see the methods for class varPower.", but I did not find the content offered in varPower to be particularly helpful in this regard. If anybody has gone through the process of creating a new varFunc I'd love to hear from you. The variance function I'd like to construct is from Bailey 1964: (m*exp((lambda-mu)*t)*((lambda/mu)+1)*(exp((lambda-mu)*t)-1)/((lambda/ mu)-1) , where m is the starting number of individuals, lambda is the birth rate, mu the death rate and t the time since the population's origin. Cheers, Rich