Predictive Models of Adult Distance Learners' Academic Performance: Comparative Analysis of Two Regression-based Models of Path Analysis

Abstract

Using two regression-based models, this study looked into factors that explain academic performance of adult distance learners. The efficiency and applicability of linear regression and logit regression procedures as statistical models for path analysis of mixed variables were compared along the criteria of model goodness-of-fit, predictive efficiency, and effect adequacy. Based on the criteria used, logit regression procedure has been shown to be a more efficient statistical tool than the classical linear regression procedure, when investigating pattern of causal relationships among variables at different levels of measurement. Implications of the study on distance education as a nontraditional mode of study and on achievement of adult distance learners were also presented.