Table 2.

Proportion of variance explained by each model.

Model	Variables	Percentage of model 1 area level variance explained
All
1	Constant, age, age-squared, age-cubed, gender and cohort	n/a
2	As model 1 adding depcat, and depcat×agea	72.2
3	As model 2 adding class, class×age, class×age-squared, class×age-cubed	84.6
4	As model 3 without depcat or depcat×age	44.1
Males
1	Constant, age, age-squared, age-cubed, and cohort	n/a
2	As model 1 adding depcat, and depcat×agea	93.1
3	As model 2 adding class, class×age, class×age-squared, class×age-cubed	100.0
4	As model 3 without depcat or depcat×age	59.7
Females
1	Constant, age, age-squared, age-cubed, and cohort	n/a
2	As model 1 adding depcat, and depcat×agea	73.9
3	As model 2 adding class, class×age, class×age-squared, class×age-cubed	84.7
4	As model 3 without depcat or depcat×age	46.8

^aInteractions between depcat and age-squared or age-cubed were left out as they were not significant at the p<0.05 level in most models. For females only the interaction effect between age-squared and being in the most deprived category was significant in Model 2 but not in model 3, suggesting that this was mainly due to effects of individual class, and so this has been left out for consistency with the overall and male only models.