Reference : Revisiting the multivariate coefficient of variation for comparing electrophoretic te...
Scientific congresses and symposiums : Unpublished conference/Abstract
Human health sciences : Public health, health care sciences & services
Revisiting the multivariate coefficient of variation for comparing electrophoretic techniques in External Quality Assessment (EQA) schemes
Zhang, Lixin[Université de Liège - ULg > Département des sciences de la santé publique > Informatique médicale et biostatistique >]
Albert, Adelin[Université de Liège - ULg > Département des sciences de la santé publique > Informatique médicale et biostatistique - Département de mathématique >]
17th annual meeting of the Belgian Statistical Society (BSS 2009)
14-16 October, 2009
[en] External Quality Assessment (EQA) schemes are national programmes designed to control the analytical performance of clinical laboratories and to maintain inter-laboratory variability within acceptable limits. In this context, the type of assay technique or equipment used by the participants is an important factor to take into account. The coefficient of variation (CV) is widely used in EQA to compare the reproducibility of techniques and equipments: the lower the CV, the better the analytical performance. The recent introduction of protein electrophoresis, a laboratory test profile consisting of five fractions (albumin, α1, α2, β and γ globulins) summing up to 100%, in the EQA programme (Zhang et. al. 2008) prompted us to revisit an old problem (Reyment 1960), namely the multivariate extension of the univariate CV concept, and to suggest some amendments to existing solutions. Further, the application of the multivariate CV to techniques and methods used in the assay of protein electrophoresis raises concomitant statistical problems. First, the compositional nature of electrophoretic data requires a log-ratio one-to-one transformation from the simplex to the real space (Aitchison 1982). Next, robust estimations of the mean and the covariance matrix (Rousseeuw and Van Driessen 1999) are needed to avoid the effect of outliers. The resulting robust multivariate CVs are illustrated on electrophoretic EQA material from the Belgian and French national EQA programmes to assess the analytical variability of the electrophoretic techniques used.