Near-constancy of the Harris function in the simple branching process

Near-constancy of the Harris function in the simple branching process

0.00 Avg rating0 Votes
Article ID: iaor19941243
Country: United States
Volume: 9
Issue: 3
Start Page Number: 435
End Page Number: 444
Publication Date: Jul 1993
Journal: Stochastic Models
Authors: ,
Keywords: branching process
Abstract:

In the theory of the simple supercritical branching process there are difficult problems concerning constancy and near-constancy phenomena. One classical case arises when the offspring generating function is polynomial of degree d. Suppose the mean family size is μ, and let γ be given by μ’γ=d. Let W be the almost sure limit of the population size normalized by its expected value. Harris established that the cumulant generating function of W, logEesW, looks like s’γH(s) for large s. (H is the Harris function of the title and is multiplicatively periodic, that is h(μs)=h(s) or all s>0.) He reports the computation of H for the offspring generating function 0.4s+0.6s2 and finds it to be constant to 6 decimal places. Here the theory necessary for a more refined numerical study is developed and the results of that study reported. The main conclusion of the numerical study is that the Harris function is not constant (at least in this case) and that, on a suitable scale, its variation is nearly sinusoidal.

Reviews

Required fields are marked *. Your email address will not be published.