Técnicas gráficas de diagnóstico para los modelos lineales mixtos
No Thumbnail Available
Date
2022-04-25
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Description
Los modelos lineales mixtos son los más utilizados en la actualidad para analizar datos longitudinales por su flexibilidad para modelar las múltiples fuentes de variabilidad presentes en este tipo de datos. Esta flexibilidad hace que el proceso de construcción del modelo no sea una tarea sencilla. Éste proceso está compuesto por reiterados ajustes y evaluaciones de modelos con el objetivo de identificar una forma funcional que describa adecuadamente la evolución de la respuesta en el tiempo, su relación con covariables y la correlación presente entre los datos. Durante este proceso, se pueden utilizar diversas herramientas diagnósticas basadas en los residuos. Éstas no solo alertan de una posible falta de ajuste del modelo a los datos, sino también, brindan información acerca de las características de los mismos, que pueden incorporarse en el modelo. Este trabajo presenta un compendio de técnicas diagnósticas gráficas basadas en los residuos y describe su utilización en el proceso de construcción del modelo y en la evaluación de los supuestos distribucionales del mismo.
Linear mixed models are currently the most widely used to analyze longitudinal data due to their flexibility to model the multiple sources of variability present in this type of data. This flexibility makes the model building process no easy task. This process is made up of repeated adjustments and evaluations of models in order to identify a functional form that adequately describes the evolution of the response over time, its relationship with covariates and the correlation present between the data. During this process, various residual-based diagnostic tools can be used. These not only alert to a possible lack of fit of the model to the data, but also provide information about the characteristics of the data, which can be incorporated into the model. This paper presents a compendium of graphical diagnostic techniques based on residuals and describes their use in the model construction process and in the evaluation of its distributional assumptions
Linear mixed models are currently the most widely used to analyze longitudinal data due to their flexibility to model the multiple sources of variability present in this type of data. This flexibility makes the model building process no easy task. This process is made up of repeated adjustments and evaluations of models in order to identify a functional form that adequately describes the evolution of the response over time, its relationship with covariates and the correlation present between the data. During this process, various residual-based diagnostic tools can be used. These not only alert to a possible lack of fit of the model to the data, but also provide information about the characteristics of the data, which can be incorporated into the model. This paper presents a compendium of graphical diagnostic techniques based on residuals and describes their use in the model construction process and in the evaluation of its distributional assumptions
Keywords
Datos longitudinales, Modelo lineal mixto, Análisis de residuos, Longitudinal data, Lineal mixed model, Residual analysis