Title: Detection and recovery from inconsistencies in the general linear model with singular dispersion matrix
Authors: Marc Hofmann - University of Oviedo (Spain) [presenting]
Ana Colubi - Kings College London (Cyprus)
Erricos Kontoghiorghes - Cyprus University of Technology and Birkbeck University of London, UK (Cyprus)
Abstract: A new method to recover from an insconsistent GLM is proposed. The GLM is reformulated as a GLLSP. The minimal set of observations that explain the inconsistencies in the model can be identified by solving a combinatorial sparse approximation problem. An exhaustive algorithm is proposed. Gram-Schmidt orthogonalization is used as the main computational tool. When the number of observations is large, non-exhaustive algorithms can be employed instead.