Results of the longitudinal study
The results of the longitudinal study showed that some of the computer tests gave a highly satisfactory prediction of students who later were found to be experiencing literacy difficulties and dyslexia. These computer tests produced significant correlations with reading development, many of which had higher correlation coefficients than were found between intelligence (verbal and non-verbal) and reading development. Over 90% of students who subsequently were found to be experiencing significant literacy difficulties were successfully predicted by the computer tests alone on school entry. Since CoPS is designed to be used as a screening device it is important to specify the levels of false negatives and false positives; CoPS produced 16.7% false negatives and 2.3% false positives. This compares very favourably with other screening devices (Singleton, 1997a). ‘False negatives’ and ‘false positives’ are the two types of classification error in screening. False negatives are cases where the screening device fails to identify a risk when a risk does in fact exist; false positives are cases where the screening device has identified a risk when a risk does not in fact exist. Various statistical techniques were used to determine which of the computer tests were most effective in predicting later difficulties (predictive validity), and eight of these were selected for the final software suite. For details of the statistical analyses carried out see Singleton, Thomas and Horne (2000).