Sobre el uso adecuado del coeficiente de correlación de Pearson: verificación de supuestos mediante un ejemplo aplicado a las Ciencias de la Salud
Cargando...
Fecha
2018
Título de la revista
ISSN de la revista
Título del volumen
Editor
Sociedad Venezolana de Farmacología Clínica y Terapéutica
Resumen
La comprobación de los supuestos en los que se sustenta el uso del coeficiente de correlación de Pearson suele ser una tarea en la que se cometen no pocos errores. Si bien es sencillo el proceso que lleva a su cálculo e interpretación, no resulta tan fácil la labor de verificar el cumplimiento de condiciones como la normalidad bivariada o la ausencia de datos atípicos, probablemente porque esto demanda la implementación de técnicas multivariantes. La presente revisión pretende servir de orientación a investigadores de las ciencias de salud y afines, los cuales seguramente se toparán con situaciones en las que deba emplearse esta herramienta estadística. El artículo gira en torno a un caso práctico derivado de un estudio de prevalencia de síndrome metabólico realizado en la ciudad de Maracaibo, Venezuela. El objetivo principal es el de mostrar mediante este ejemplo la manera adecuada de constatar las premisas vinculadas a este coeficiente, no sin olvidar el debido argumento teórico que las respalda. Se prescinde del aspecto matemático en favor del informático, para lo cual se utiliza el programa abierto R-Studio en todas y cada una de las actividades de procesamiento, diagramación y cómputo. Se proveen las bases de datos empleadas en el desarrollo del problema, a la vez de suministrar los scripts que activan las funciones del paquete con el propósito de que el lector pueda reproducir el análisis y comparar los resultados. Toda esta información puede ser consultada y descargada desde un repositorio de libre acceso.
The checking of the assumptions on which the use of the Pearson correlation coefficient is based, is usually a task in which many errors are committed. Although the process that leads to its calculation and interpretation is simple, the task of verifying conditions such as bivariate normality or the absence of outlier is not so easy, probably because this requires the implementation of multivariate techniques. This review intends to serve as guidance to health sciences researchers, who will surely find situations in which this statistical tool should be used. The article is based on a prevalence study of metabolic syndrome carried out in the Maracaibo city, Venezuela. The main objective is to show by this example the appropriate way to verify the assumptions linked to this coefficient, not forgetting the due theoretical argument that supports them. The mathematical aspect is discarded in order to get the benefits of using computers power, for which the open source R-Studio program is used in each and every one of the processing, plotting and computation activities. The dataset used in the development of the problem are provided, as well as the scripts that activate the functions of the package with the purpose that the reader can reproduce the analysis and compare the results. All this information can be consulted and downloaded from an open access repository.
The checking of the assumptions on which the use of the Pearson correlation coefficient is based, is usually a task in which many errors are committed. Although the process that leads to its calculation and interpretation is simple, the task of verifying conditions such as bivariate normality or the absence of outlier is not so easy, probably because this requires the implementation of multivariate techniques. This review intends to serve as guidance to health sciences researchers, who will surely find situations in which this statistical tool should be used. The article is based on a prevalence study of metabolic syndrome carried out in the Maracaibo city, Venezuela. The main objective is to show by this example the appropriate way to verify the assumptions linked to this coefficient, not forgetting the due theoretical argument that supports them. The mathematical aspect is discarded in order to get the benefits of using computers power, for which the open source R-Studio program is used in each and every one of the processing, plotting and computation activities. The dataset used in the development of the problem are provided, as well as the scripts that activate the functions of the package with the purpose that the reader can reproduce the analysis and compare the results. All this information can be consulted and downloaded from an open access repository.
Descripción
Palabras clave
Coeficiente de correlación, Pearson, Supuestos, R-Studio, Caso práctico, Síndrome metabólico, Maracaibo, Correlation coefficient, Pearson, Assumptions, R-Studio, Practical case, Metabolic, Syndrome