Implications of raw score and regression based definitions of reading disability

Consequences of the variations in the type of criteria used to define reading disability were examined in 1069 9-to-14 year old children. Two definitions were used to select impaired readers: 1) a raw score discrepancy between WRAT Reading scaled score and WISC Full Scale IQ and 2) a discrepancy of one standard error between actual WRAT Reading scaled score and that score predicted on the basis of WISC Full Scale IQ by regression procedures. Outcome neuropsychological measures included ten tests selected from the Halstead Reitan Battery for Children (Rourke, Bakker, Fisk & Strang, 1983). Using a raw score discrepancy definition lead to an artificial elevation in IQ in the disabled group and subsequent regression artifacts. Initial group differences were small in magnitude, but were subsequently enhanced by controlling for IQ. Regression based definition does not yield an IQ elevation in the disabled readers because groups are statistically equated with respect to IQ. Correspondingly, the enhancement in effect size due in response to controlling for IQ was smaller when applied to groups defined by regression based criteria. Comparisons of the joint effects of the two types of definition were made by identifying three groups of disabled children: those who were impaired under both raw score and regression based definitions (IBOTH) and those who met only one definition; either raw score (IRAW) or regression based (IREG) selection criteria. Results revealed that all three reading groups differed significantly in cognitive skill from each other. Conversely, when effects of IQ were removed, group differences remained only between the IBOTH and IRAW reading groups and these were largely verbal in nature.