Models of Competence and Data Mining

Although we know there is a relationship between OTL and student achievement, and such a relationship is multidimensional and domain-specific, a natural question to ask is how OTL is related to student competence. Because competence is categorical, i.e., reached or not reached, in order to differentiate the relationship between two continuous variables (e.g., per pupil expenditure and student science achievement scores) from the relationship between categorical variables (e.g., competence and teaching practices), I call the relationship between OTL and competence models of competence, or competence models.

Many methods are available to develop models of competence. If models of competence assume a causal relationship between OTL and competence, randomized experimental designs or their approximations (such as quasi-experimental designs) are ideal but unfeasible. For example, withholding OTL from some students is unethical and should never be done in educational research. The unfeasibility of experimental studies leaves the methodological options to statistical methods, especially causal comparative methods. Causal comparative methods are based on correlations, supplemented with conceptual justification for causality. Examples of such statistical methods based on correlations are logistic regression, discriminant analysis, and profile analysis.