Estimating the concordance correlation coefficient with skewed data

The Concordance Correlation Coefficient has been one of the key analysis measures of concordance in particular for the repeated measures outcomes in the continuous scale. The coefficient relies on a key distributional assumption of normality of the response variable. In this thesis the main estimati...

Descripción completa

Detalles Bibliográficos
Autor: Peón Pena, Gonzalo
Tipo de recurso: tesis de maestría
Fecha de publicación:2025
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/424632
Acceso en línea:https://hdl.handle.net/2117/424632
Access Level:acceso abierto
Palabra clave:Linear models (Statistics)
Mathematical statistics
Bootstrap (Statistics)
Concordance
Models lineals (Estadística)
Estadística matemàtica
Bootstrap (Estadística)
Classificació AMS::62 Statistics::62F Parametric inference
Classificació AMS::62 Statistics::62G Nonparametric inference
Classificació AMS::62 Statistics::62J Linear inference, regression
Àrees temàtiques de la UPC::Matemàtiques i estadística
Descripción
Sumario:The Concordance Correlation Coefficient has been one of the key analysis measures of concordance in particular for the repeated measures outcomes in the continuous scale. The coefficient relies on a key distributional assumption of normality of the response variable. In this thesis the main estimation approaches for the longitudinal CCC have been reviewed through simulation sets under distributional misspecifications, where both bayesian and bootstrap Bca approaches obtained the better coverage, but nonetheless all reviewed methods failed to reach its nominal coverage under strong right-skewness on the response variable. The application of popular transformations for the asymptotic estimate did not provide visible coverage improvements, while in the presence of a drop-out pattern robust methods that require complete case analysis were significantly limited for the longitudinal CCC estimation.