Authors' reply

We welcome the opportunity to clarify the points raised by D. Marchevsky.

1. (a) The non-clinical sample was selected by asking the head teachers of each school to choose a selection of mixed-ability classes from each of the school year groups that fitted the age range we had selected.

2. (b) Of the 180 adolescents re-tested, 77 (43%) were girls, almost identical to the ratio of the first test sample (44%). The variances in the different groups are indeed heterogeneous, but the results of the analysis hold when the analyses are corrected for this effect or when non-parametric analyses are carried out. Limited space precluded us from reporting the full analyses.

3. (c) The results remain the same whether or not the deliberate self-harm group is included in the analyses.

4. (d) Robust multiple comparisons show that, for the depression sub-scale, the out-patients depressed group scores higher than the other three groups, with the other two clinical groups not differing significantly from each other, but both scoring significantly higher than the school sample. For the anxiety sub-scale, the three clinical groups do not differ significantly from each other, but all score significantly higher than the school sample.

5. (e) Analyses for each gender separately produce the same results.

6. (f) We see no problems with using the Pearson product-moment correlation as a measure of test-retest reliability. There is a long history of using this correlation as a measure of reliability in the psychometric test theory literature. Note that we are not measuring agreement between raters here, for which a measure such as kappa would be appropriate.