# Table 6 School effects and models to explain these effects (the models are those described fully in Table 3)1

Boys' Girls'
Between-school variance2 (standard error) 95% credible intervals 3 & 4 % of model 1 variance5 Between-school variance (standard error) 95% credible intervals4 % of model 1 variance
Model 1 – adjusted for socio-demographic and cultural factors predicting smoking. 0.250 (0.099) (0.114,0.497) (100%) 0.046 (0.021) (0.012,0.081) (100%)
Model 2 – adds (to model 1) individual cognition measures. 0.275 (0.111) (0.122,0.551) (110%) 0.048 (0.023) (0.015,0.112) (104%)
Model 3 – adds (to model 2) cognitions relating to school. 0.240 (0.105) (0.116,0.520) (96%) 0.030 (0.014) (0.011.0.067) (65%)
Model 4 – adds (to model 3) school level affluence (factor 1). 0.175 (0.077) (0.068,0.302) (70%) 0.040 (0.024) (0.011,0.101) (86%)
Model 5 – adds (to model 3) school level poor quality of relationships – factor 3 0.201 (0.084) (0.088,0.408) (80%) 0.038 (0.021) (0.011,0.91) (83%)
Model 6 – adds (to model 5) school level affluence (factor 1), and the interaction between factor 1 and poor quality of relationships. 0.045 (0.023) (-0.001,0.112) (18%) 0.039 (0.020) (0.000,0.077) (85%)
1. 1Statistically significant results are bolded
2. 2 Between-school variance is the difference between the average value of smoking at school level (based on the data i.e. actual smoking of pupils in a school) and the estimates obtained from the modelling. The aim is to develop models whereby the actual school level smoking values are close to the estimated values, thus lowering the variance is what is desired.
3. 3 The Credible Intervals (produced as part for the MCMC estimation procedure) represent the difference in an outcome between a school at the bottom (2.5th centile) and the top (97.5th centile) of the distribution.
4. 4 Results are significant at the alpha = 0.05 level when the 95% credible intervals for the between school variance do not include zero.
5. 5 As described in the Introduction, a 'school effect' is when 'pupil outcomes for a school vary, either positively or negatively, from that which might be expected, given the known predictors of these outcomes' (i.e. between-school variance after adjusting for Model 1 predictors). Given that Model 1 adjusts for the known predictors of smoking, the results for Model 1 indicate that there is a significant 'school effect' on males' and females' smoking behaviour at age 16. Model 6 has successfully explained that 'school effect' for males, while Model 3 explained the 'school effect' for girls.