Abstract
For many countries, in particular in sub-Saharan Africa, Demographic and Health Surveys (DHS) are the main estimates of HIV prevalence at national levels in general population. Several DHS collect longitude and latitude of surveyed clusters.In this paper, we present three methodological approaches for mapping spatial variations of HIV prevalence from DHS. These approaches are applied to DHS simulation sampled from a model country. The estimated surfaces are then compared with the initial surface of the model.We show that a method using kernel estimators with adaptive bandwidths of the same number of observed people allows estimating main regional trends of the epidemics. Its application to data from 2003 DHS of Burkina Faso give a plausible picture of the epidemiological situation in this country.