A SEMI-MARKOV MODEL FOR INSECT LIFE-HISTORY DATA

We describe a semi-Markov model for an insect's life history, the motivation of which is biologically appealing. We assume that development follows a Brownian motion process with drift. Stage occupancy is determined on the basis of the stage recruitment times, which are inverse Gaussian random variables. Independent of development, mortality is assumed to be a two-state nonhomogeneous Markov process. The resulting model is applicable for data arising from single-generation populations composed of single or multiple cohorts of insects. The model is fitted to stage-frequency data assuming a product-Poisson likelihood for the counts. Maximum likelihood estimates of the parameters and their asymptotic standard errors are obtained via a Gauss-Newton algorithm for iteratively reweighted least squares. An example is provided, which illustrates some of the features of both the model and the data themselves.