Coeficiente de concordancia para variables continuas
Fecha
2010-03-09
Autores
Camacho Sandoval, Jorge
Título de la revista
ISSN de la revista
Título del volumen
Editor
Colegio de Médicos y Cirujanos, Costa Rica
Resumen
En las notas estadísticas anteriores, se hizo referencia a medidas de asociación entre dos variables. En el caso particular de dos variables continuas, se estudió el coeficiente de correlación de Pearson y su forma de cálculo. Una situación diferente surge cuando se desea medir la misma variable, en las mismas muestras o pacientes, con dos métodos, equipos o personas diferentes, para determinar si ambos métodos, equipos o personas producen resultados equivalentes. En ese caso lo que interesa es determinar si ambas mediciones son similares en magnitud, no si están asociadas, de hecho deben estarlo, ya que son dos mediciones de la misma característica en los mismos individuos o muestras. No obstante, si una de las mediciones tiene un error sistemático, por ejemplo si una de las mediciones tiene sistemáticamente cinco unidades menos que la otra medición, el coeficiente de correlación puede ser muy elevado aunque las diferencias en las mediciones sean importantes, es decir, las mediciones pueden no ser concordantes.
In the previous statistical notes, reference was made to measures of association between two variables. In the particular case of two continuous variables, the Pearson correlation coefficient and its calculation method were studied. A different situation arises when you want to measure the same variable, in the same samples or patients, with two different methods, equipment or people, to determine if both methods, equipment or people produce equivalent results. In that case, what is interesting is to determine if both measurements are similar in magnitude, not if they are associated, in fact they must be, since they are two measurements of the same characteristic in the same individuals or samples. However, if one of the measurements has a systematic error, for example if one of the measurements is systematically five units less than the other measurement, the correlation coefficient can be very high even though the differences in the measurements are important, that is, measurements may not be consistent.
In the previous statistical notes, reference was made to measures of association between two variables. In the particular case of two continuous variables, the Pearson correlation coefficient and its calculation method were studied. A different situation arises when you want to measure the same variable, in the same samples or patients, with two different methods, equipment or people, to determine if both methods, equipment or people produce equivalent results. In that case, what is interesting is to determine if both measurements are similar in magnitude, not if they are associated, in fact they must be, since they are two measurements of the same characteristic in the same individuals or samples. However, if one of the measurements has a systematic error, for example if one of the measurements is systematically five units less than the other measurement, the correlation coefficient can be very high even though the differences in the measurements are important, that is, measurements may not be consistent.
Descripción
Nota estadística de un artículo científico dividido por partes, realizado por Jorge Camacho Sandoval
Palabras clave
ESTADÍSTICA, INVESTIGACIÓN, VARIABLES, INVESTIGACIÓN CUALITATIVA, MÉTODOS ANALÍTICOS, STATISTICS, INVESTIGATION, VARIABLES, QUALITATIVE RESEARCH, ANALYTICAL METHODS