, 2008). In order to test the need for cross-classification by neighbourhood (LSOA),

models with and without neighbourhood cross-classification were tested at this stage. The ranking of schools based upon the extent to which the observed mean BMI-SDS differed from the expected mean BMI-SDS was recorded (Expected residuals). Schools with observed mean pupil weight status which is markedly different from that expected (i.e. high or low residuals) may represent hot and cold spots of obesity. Calculate and rank schools according to a ‘value-added’ score (‘Value-added’ ranking) The ‘Expected’ ranking gives a measure of the impact of the school, but does not account Ibrutinib for pre-school weight status. As the data were cross-sectional, differences within-pupils could not be calculated.

Instead, differences between year groups of pupils were calculated through an identical process to that used by Procter et al. (2008). As Reception is the first year of schooling Reception pupils are relatively unexposed to the school environment and context compared with pupils in Year 6, and therefore the Reception pupil weight status was conceptualised as the pre-school weight status. The expected residuals for Reception and Year 6 pupils were calculated separately using the same multilevel model as in Step 2. The difference between these two sets of expected residuals gave a http://www.selleckchem.com/products/Y-27632.html measure (score) of the average ‘value-added’

to the pupil BMI-SDS by the school, the ranking of which was recorded. Compare the Observed, ‘Expected’ and ‘Value-added’ rankings. Primarily Lin’s concordance correlation coefficients (ρc) ( Lin, 1989, Lin, 2000 and Steichen and Cox, 2002) were used to quantify the agreement between pairs of rankings within each of the five years. Pearson’s correlation coefficients (r) were calculated alongside the concordance values, and the Ketanserin rankings were visualised in caterpillar plots; these additional analyses are reposted in the supplementary material. Compare stability of the rankings across the five years (2006/07–2010/11) Within each ranking, concordance correlation coefficients were calculated comparing the agreement between each of the five years of rankings. As with the previous step Pearson’s correlation coefficients and caterpillar plots are reported as supplementary material. Tracking coefficients (kappa) were calculated to explore the extent to which schools maintained approximately the same rankings across the five years. In order to quantify approximate positions, the rankings of schools were split into quintiles each year, prior to the calculation of the tracking coefficients. There was no comparison between the three types of ranking in this step.