External validity and the choice of statistical method: research in reading