Skip to main content

Table 3 Sequences of models tested

From: What explains between-school differences in rates of smoking?

Model 1 This model controlled for socio-demographic and cultural factors predicting smoking. Any 'school effect' can only be identified after adjustment for such factors.
  The variables included were; pupils' gender, age in months at follow-up, cohort, whether left school at youngest possible legal age, composition of the family unit in which pupils lived, housing tenure, levels of parental monitoring, father's social class, amount of personal spending money, ethnic group, and mother's age. Only one interaction with gender was found and added to the model – gender and living with mother only. All PQ data.
Model 2 This model adds (to Model 1) measures of self esteem and religiosity that may be influenced by school (but are likely to be more influenced by home environment). All PQ data.
Model 3 This model adds (to Model 2) cognitions related to school, that is, pupils' attitude to school and pupil-assessed teacher-pupil relationships. It was expected that these cognitions would be influenced by school experience (though, of course, home environment may also have been important). All PQ data
Model 4 This model adds (to Model 3) school level affluence (factor 1 – LEA data).
Model 5 This model adds (to Model 3) teacher-assessed school-level (poor) quality of relationships (both staff-staff and staff-pupil relationships – factor 3). All TQ data.
Model 6 This Model adds to (Model 5) school level affluence (factor 1 – LEA data), and the interaction between factor 1 and poor quality of relationships.