Bias Correction in Clustered Underreported Data

Data quality from poor and socially deprived regions have given rise to many statistical challenges. One of them is the underreporting vital events leading biased estimates for associated risks. To deal with underreported count data, models based on compound Poisson distributions been commonly assumed. be identifiable, such usually require extra strong information about probability reporting event in all areas interest, which not always available. We introduce a novel approach model assuming that are clustered according their data quality. leverage these clusters create hierarchical structure probabilities decrease as we move best group worst ones. obtain constraints identifiability prove only prior experiencing required. Several approaches uncertainty presented, including reference priors. Different features regarding proposed methodology studied through simulation. apply our map early neonatal mortality risks Minas Gerais, Brazilian state presents heterogeneous characteristics relevant socio-economical inequality.