Model-cum-design-based estimation of the prevalence rate of a disease in a locality using spatial smoothing

Abstract

In this article, we study the problem of estimating the prevalence rate of a disease in a geographical area, based on data collected from a sample of locations within this area. If there are several locations with zero incidence of the disease, the usual estimators are not suitable and so we develop a new estimator, together with an unbiased estimator of its variance, which may be appropriately used in such situations. An application of this estimator is illustrated with data from a large-scale survey, which was carried out in the city of Kolkata, India, to estimate the prevalence rate of stroke. We show that spatial modelling may be used to smooth the observed data before applying our proposed estimator. Our computations show that this smoothing helps to reduce the coefficient of variation and such a model-cum-design-based procedure is useful for estimating the prevalence rate. This method may of course be used in other similar situations.

Keywords:

• conditional autoregressive model
• Hartley–Ross estimator
• spatial dependence

Acknowledgements

This study and the survey referred to here were funded by a grant from the Indian Council of Medical Research. We are grateful to two referees for their constructive and insightful comments which led to improvements over an earlier manuscript.