Skip to main content

Applications of the gibbs sampling algorithm to multiple outlier detection in econometric models - the masking problem

Objective



The proposal of this project is the research and development of new methods to identify groups of outliers in econometric models. When multiple outliers exist, the standard detection statistics for a single outlier are not reliable: they may not identify groups of outliers and they may identify as outliers data that are not. This is the masking problem. The new procedures will be develop from the Bayesian point of view and using Markov Chain Monte Carlo (MCMC) methods.

Coordinator

UNIVERSITE CATHOLIQUE DE LOUVAIN
Address
34,Voie Du Roman Pays 34
1348 Louvain-la-neuve
Belgium

Participants (1)

Not available
Spain