Spurious statistics 1

The Cranfield study of correlations between 2001 RAE scores against citation counts ("REF will topple RAE stars, report warns", 14 February) is less valuable than it could have been, as it was based on a calculation of average number of citations per article examined. All previous studies (which have shown excellent correlations between RAE scores and citation studies and therefore indicate citation analysis would not greatly affect overall order of RAE score) instead examined total number of citations per department or average number of citations per member of staff. Perhaps the Cranfield authors should be asked to re-run their study following this standard methodology?

Charles Oppenheim
Head of the department of information science
Loughborough University