We present a technique for estimating the size of a statistical population spread across multiple sources. Each sources can overlap others, possibly leading to over-counting individuals in naive approaches. The methodology we introduce further guarantees that the process is performed in an anony- mous way, i.e. without explicitly sharing individual data. We propose to apply this approach to the medical sector, where the input data are particularly sensible, and where anonymity is usually a requirement. Our methodology applies well for coordinated counting estimates, but can be in principle extended to generate data sets.
Voir en ligne : http://www.icts4ehealth.icar.cnr.it/