Confounding the targets;Letter
Educationists who place value on the maintenance of standards should never accept the current use of the free school meal (FSM) method because it contains mistakes in statistical inferencing of the type that would be heavily penalised even at undergraduate level.
While researchers agree that there is a strong (negative), but far from perfect, correlation between FSM entitlement and Standard grade performance, the Audit Unit goes on, in effect, to draw the invalid conclusion that differences in Standard grade results between schools with similar FSM can be largely and reliably explained by differences in the quality of teaching and management.
What the Audit Unit has not taken seriously enough is that research has shown that there is a very strong (positive) correlation between success at Standard grade (especially at Credit level) and having parents with high educational qualifications and high occupational status.
Even cursory examination of the data and a modicum of knowledge of the schools and communities they serve should have aroused strong suspicions that schools with similar FSMs frequently have significantly different proportions of pupils from educational advantaged homes.
Failure to adequately control or statistically adjust for such a factor is a serious and fundamental mistake in project design and data interpretation. An inadequately controlled factor such as this acts as a confounder, meaning that, in the case of any two schools of similar FSM, it is not possible to separate the influence of the confounding factor from that of school quality factors in accounting for any difference in Credit level Standard grade results between the two schools.
Because this confounding effect runs through all the data, the Audit Unit's benchmarking smoothing exercise, in which schools obviously anomalous in terms of size and Standard grade results are "set aside", does not cure the problem. Nor does allowing local authorities to make minor adjustments to targets on the basis of no agreed criteria.
The Audit Unit's method come nowhere near being reliable enough for the purposes of differential target-setting.
Michael Davenport Bank Road, East Linton, East Lothian