Sarojinie Fernando

IFS/Statistics, Massey University, Palmerston North

Estimation of geographical relative risk by local polynomial regression

A common problem in epidemiology is the estimation of spatial variation in risk associated with a disease. In this talk, we focus our attention on the estimation of a relative risk function using data on the locations of cases and controls in geographical epidemiology. This can be obtained by using a ratio of two bivariate kernel density estimates (the density ratio method). A novel alternative is to use local polynomial regression to estimate the relative risk function. We have studied the properties of the latter method and derived the asymptotic bias and variance of local polynomial estimates. The performance of estimators based on the above mentioned methods are compared through a Monte Carlo simulation study and illustrated using five synthetic problems. An application on disease in myrtle trees is presented.