CONFIGURAL FREQUENCY EVALUATION OF DATA F ROM PRETEST-POSTTEST CONTROL-GROUP DESIGNS

For treatment evaluation in multivariate pretest-posttest designs, met hods of Configural Frequency Analysis (CFA) are recommended. Specifica lly, three CFA methods are proposed: (1) Prediction CFA, (2) Interacti on CFA, and (3) Generalized Unary Hypothesis Analysis (GUHA) CFA. Thes e methods share in common that they define treatment response-relation s via so-called types, that is local or regional associations among th e factors of a treatment-response cross-classification. Unit of analys is for these CFA approaches are difference scores between the first an d second observation. Numerical examples use data from an experiment o n the effects of training methods on calculation and reasoning in lear ning disabled school children.