Q2. What are the future works in "Predicting small-area health-related behaviour: a comparison of smoking and drinking indicators" ?
Further work has explored the relationship between these ` synthetic estimations ' and the results of local surveys and the impact of using a multilevel approach ( Moon et al., 1998 ). For the purposes of this methodological paper however, two key concluding points can be made.
Q3. How much does a single district health authority cost?
Well-structured surveys with sound sampling design capable of generating representative results at the subdistrict level are estimated to cost at least £50,000 for a single district health authority.
Q4. What is the effect of cross-level interactions on smoking?
Cross-level interactions suggest that single women living in areas with high percentages of private rented households have a raised likelihood of problem drink-ing while single women living in a uent areas are more likely to smoke.
Q5. What are the key contextual variables in the UK Census?
Other important contextual variables such as tenure andsocial class are not available in a small area cross-tabulation, which also contains age and gender thus losing the basis for age-sex standardisation.
Q6. Why is it important to include the level of household?
Additionally there are also substantive reasons for not including the household level: in a correct model, the overall e ect of including the level of household would be to reduce variation at the higher levels.
Q7. What is the effect of the surrogate of dual car ownership?
the a uence of an area, as measured by the surrogate of dual car ownership, equates with a reduced likelihood of smoking but an increased likelihood of problem drinking.
Q8. What are the two factors that are particularly important in the case of problem drinking?
Male gender and being single are particularly important factors in both cases, the former notably so in the case of problem drinking.
Q9. What is the main reason for the disaggregation of national surveys?
In the main however the sampling design of national surveys is insu ciently robust to permit disaggregation below the scale of 14±25 regions Ð and, paradoxically, any attempts at such disaggregation could only be validated with local survey data.
Q10. What is the detailed crosstabulation available at the census ward level?
The most detailed such crosstabulation available at the census ward level and relevant to health-related behaviour is age (grouped into several age bands), marital status and gender.
Q11. Why were the parameters chosen as appropriate measures to achieve standardisation of individual responses and capture the impact?
the chosen parameters were theoretically justi®ed as appropriate measures to achieve standardisation of individual responses and capture the impact of deprivation ecologies and, while more complex models might have been developed had the objective of the study been the description of health-related behaviours, few parameters were used because of the limits imposed by the need subsequently to use the Census to generate predictions.
Q12. What was the first strategy for identifying local variations in health-related behaviour?
In the late 1980s, geodemographics emerged as a potential third strategy for identifying local variations in health-related behaviour (Speller and Hale, 1985).
Q13. What is the importance of the census as a basis for local predictions of health-related behaviour?
The importance of the census as the basis for local predictions of health-related behaviour also constrained the individual-level explanatory variables.