Correlaciones múltiples y correlaciones canónicas en la predicción de rendimiento académico

Abstract

A sample of Psychology students answered the Kuder preference scales, the verbal reasoning, numeric abilities, abstract reasoning, Space relationships, spelling and phrases sub-tests of the DA T and the Maudsley Personality inventory. When they finished their first year of school their grades were correlates and a regression equation was calculated. The seven variables which best predicted achievement were: the areas of scientific interest, persuasive interest, musical interest and social service; as well as verbal reasoning, abstract reasoning and spelling. Four years later the total GPA and the areas GPA was calculated for those who did graduate. Canonical correlations were calculated for the seven best predictors and the subject areas as well as a multiple correlation with the general GPA. From the results one can conclude that although the first two canonical correlations were greater than any of the correlation coefficients between the original variables, these values are not significantly different from zero; the multiple correlation coefficient was significant. Two of the predictor variables (musical interest and abstract reasoning) did not correlate significantly with any of the criteria and the regression coefficients are practically zero. The best predictors were persuasive interest, social service and the spelling sub-test.