A recently proposed procedure for correcting inconsistent (i.e. incoherent) probability assessments is specifically tailored for the statistical matching problem with misclassification component. Such procedure is based on L1 distance minimization encoded in mixed integer programming (MIP) problems and it results particularly apt to deal with assessments stemming from different sources of information. The statistical matching problem is one of those cases. The statistical matching problem has been recently studied also inside a misclassification setting. To proceed with a correction in such a framework, if marginal assessments on the conditioning event are wanted to remain fixed, the only possible solutions are the closest Fréchet–Hoeffding bounds for the misclassification probabilities. On the contrary, if also the marginal probabilities are allowed to be modified, the L1-based procedure can be applied by a straightforward translation in an MIP problem. Such procedure is applied to a healthcare expenditures and health conditions data example.
Probabilistic inconsistency correction for misclassification in statistical matching, with an example in health care
Capotorti A.
Membro del Collaboration Group
2020
Abstract
A recently proposed procedure for correcting inconsistent (i.e. incoherent) probability assessments is specifically tailored for the statistical matching problem with misclassification component. Such procedure is based on L1 distance minimization encoded in mixed integer programming (MIP) problems and it results particularly apt to deal with assessments stemming from different sources of information. The statistical matching problem is one of those cases. The statistical matching problem has been recently studied also inside a misclassification setting. To proceed with a correction in such a framework, if marginal assessments on the conditioning event are wanted to remain fixed, the only possible solutions are the closest Fréchet–Hoeffding bounds for the misclassification probabilities. On the contrary, if also the marginal probabilities are allowed to be modified, the L1-based procedure can be applied by a straightforward translation in an MIP problem. Such procedure is applied to a healthcare expenditures and health conditions data example.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.