A Bayesian survey technique for rare subjects

We propose a methodology to draw a representative sample from a population with a small fraction of potential customers for a certain product or a service. The objective is to increase the expected number of potential customers in a sample of fixed size representing the whole population. The method also provides a useful way of incorporating subjective judgement and uncertainty into the sampling plan. This approach has produced very satisfactory results for rare event problems in a number of telecommunications applications.     

Estimation of innovation diffusion models with application to a consumer durable

A class of birth processes having a variety of practical applications in penetration of new services and products is considered. Typically, statistical inferences on these models are performed by means of simple error structures placed on the deterministic analogs of the underlying stochastic processes. Motivated by the poor performance of conventional estimation methods, the problem of estimating the parameters of these models is readdressed. We develop necessary formulae for performing the maximum likelihood estimation and weighted least squares estimation methods, and demonstrate their superiority through analyses of some real data and simulation studies.