Modelling the cause of dependency with application to filaria infection.

A preliminary data set is analysed containing filaria specific IgG4 and IgE levels and the presence of microfilariae of 196 people from families of a village in Indonesia. Since filaria infected people may not be microfilaria positive, a filaria infection can easily be missed. First, the probabilities of a filaria infection are estimated from the IgG4 levels and the presence of microfilariae using the EM algorithm. By dichotomizing these probabilities, infection status is estimated for each person. Then for IgG4, IgE and infection status, the correlations between observations are modelled. Three causes for a correlation are considered, namely genetic, intra-uterine or environmental effects. The correlation structure of the genetic and the intra-uterine effects are quite similar and consequently it may be difficult to disentangle them. Empirical variograms are plotted and the various variance components are estimated by maximizing the log-likelihood. For infection status an environmental effect is found and for IgG4 and IgE levels genetic effects are found.