A quasi-likelihood approach to model bernoulli data with spatial dependence

Abstract

A Binomial random variable is defined as a sum of n independently identically distributed binary random variables. However, departure from independence is frequently encountered in practice. This paper proposes a method to estimate one type of spatially dependent model, which uses the moments of these models and a quasi-likelihood procedure. An application in the integrated circuit manufacturing is used to illustrate the technique.

Keywords:

• dependent Bernoulli
• join count statistics
• overdispersion
• spatial log odds ratio