Relatrice
Fulvia Mecatti
Università di Milano Bicocca
Abstract
The idea of improving the quality of sample data by shaping the collection upon specic, challanging features of the study, is an old one. In observational studies it can be traced since the 40's, with the introduction of combinations of design & estimator unbiased despite being based on data that have been collected failing to ensure equal selection probabilities to all population units. Moreover, as in the words of S.K.Thompson at the 25th Morris Hansen Lecture, besides the two traditional uses of population-based sampling designs, namely making inference on population quantities and setting experiments into population, a third objective is receiving increasing attention: making interventions to change populations, which can hardly be pursued by means of traditional equal probability designs. In this perspective an emblematic example in the bio-medical eld are large scale surveys sponsored by WHO for assessing tubercolosis prevalence at national level. Inspired by this example a proposal will be illustrated for a sampling strategy aiming at improving the eectiveness of resource-consuming surveys of the sort. It is based on tailoring the data collection by combining an adaptive approach with a sequential selection. The adaptive component has the purpose to improve the positive-case detection rate, by exploiting the spatial clusterization typical of an infectuous disease. The addition of a sequential component aims at providing a exible formal framework for dealing with budget and logistic constraints, while at the same time fostering the case-oversampling goal. Empirical evidence, via simulation, will be given for discussing relevant practical matters such as the control over the sample size and pro's and cons over traditional sampling strategies.
Organizzazione
Daniela Cocchi