Amenazas a la validez analítica de las técnicas usadas habitualmente en la evaluación de programas
Loading...
Publication date
Advisors
Department
Research group
Center
Abstract
Técnicas basadas en el modelo lineal general son frecuentemente
recomendadas para el análisis de datos obtenidos
desde diseños cuya particularidad más relevante es la asignación
de grupos intactos a las condiciones investigación. A pesar
del apoyo que tradicionalmente tan recibido estos procedimientos,
con en el presente trabajo se pretende, por un lado, alertar
a los investigadores aplicados de los errores de inferencia que
pueden cometer cuando en el proceso de evaluación científica
utilizan los modelos estadísticos convencionales para establecer
la efectividad de un tratamiento administrado a grupos de
sujetos, más que sujetos individuales. Y, por otro lado, sugerirles
procedimientos analíticos válidos y fáciles de implementar
mediante alguno de los diferentes programas genéricos existentes
(p.e., mediante el módulo PROC MIXED del programa
SAS).
Techniques based on the general linear model are frequent/y recommended for the analysis of data obtained from designs whose more notable particularity is that intact groups are assigned to specific treatment conditions. Despite the advocacy received by these procedures, with this paper is sought, on one hand, to alert the applied investigators of the inferential errors that can make when in the process of scientific evaluation the usual statistical models use to found the effectíveness of an intervention administered to groups of subjects rather than to individual subjects. And, on the other hand, to suggest them va lid and easy analytic procedures of implementing by means of some of the different existent generic programs (e.g., trough the MIXED procedure of SAS program).
Techniques based on the general linear model are frequent/y recommended for the analysis of data obtained from designs whose more notable particularity is that intact groups are assigned to specific treatment conditions. Despite the advocacy received by these procedures, with this paper is sought, on one hand, to alert the applied investigators of the inferential errors that can make when in the process of scientific evaluation the usual statistical models use to found the effectíveness of an intervention administered to groups of subjects rather than to individual subjects. And, on the other hand, to suggest them va lid and easy analytic procedures of implementing by means of some of the different existent generic programs (e.g., trough the MIXED procedure of SAS program).







