Maximum likelihood estimation of the offspring variance in a Bienaymé-Galton-Watson branching process



next up previous
Next: Introduction

Maximum likelihood estimation of the offspring variance in a Bienaymé-Galton-Watson branching process

Peter Guttorp
Department of Statistics, GN-22
University of Washington
Seattle, WA 98195
USA Richard A. Lockhart
Department of Mathematics and Statistics
Simon Fraser University
Burnaby, BC V5A 1S6
Canada

Abstract:

An algorithm for the computation of a maximum likelihood estimate of the offspring distribution in a Bienaymé-Galton-Watson branching process is presented. Although the offspring distribution in general is not consistently estimable (even conditional upon non-extinction), the invariance of maximum likelihood estimators allows for estimation of the offspring variance as the variance of the estimated offspring distribution. The variance is (conditionally) consistently estimable assuming fairly strong regularity conditions. The proof uses a new local limit theorem for discrete random variables, which is uniform over distributions of different lattice size.





Richard Lockhart
Thu Oct 26 23:26:04 PDT 1995