This paper develops a class of autoregressive and moving average models
which extend the generalized linear model. Likelihood and quasi-likelihood
estimation procedures are developed which allow the models to be easily
estimated and tested. Several examples are given which illustrate the
usefulness and simplicity of the approach advocated in this paper.


KEYWORDS: ARCH, autoregression, binomial, categorical data, count
data, diagnostic checking, exponential family, gamma, generalized linear
models, martingale difference, moving average, overdispersion, poisson.