Using this data reduction procedure allowed two DFAs to be conducted on the
two kinds of classrooms. The DFAs showed that 69.6% of the teacher-centered
classrooms and 65.1% of the student-centered classrooms had been classified correctly. Two PCAs were rotated using the varimax procedure. Loadings ranged from .46 to .78 and from .63 to .81, respectively. Wilks’s lambda was calculated, showing that the discriminant model as a whole was statistically significant. Logistic regression was used in conjunction with the DFAs because it is less restrictive for studies that have unequal group sizes. There was also a second Wilks’s lambda performed to test the significance of the eigenvalue for each discriminant function. In both cases, the lambda was less than 1.00 and therefore significant, confirming that the group means were significantly different. The effect sizes, as calculated by the I index (Huberty & Lowman, 2000) were .55 and .48 for teacher-centered and student-centered classrooms, respectively.