Objective: The aim of the present paper was to give insight into the practical consequences of using either single-level or multilevel regression analyses on data from research on environmental determinants of physical activity.
Methods: For this purpose, results from single-level and multilevel regression analyses on comparable attributes of the environment were compared using a combination of individual and aggregated data, merged at the neighbourhood level.
Results: Using only individual level data, applying multilevel instead of single-level analyses did not substantially influence the results. However, ignoring the multilevel structure of our data by applying single-level in stead of multilevel analyses resulted in statistically significant associations for all the environmental attributes under study. Moreover, using information on environmental attributes at both the individual and the contextual level to adjust associations at one level for the other level showed that associated environmental attributes tend to be associated either at the individual or at the contextual level.
Conclusions: These results stress the importance for reviews and meta-analyses of recording type of measurement and type of analytical strategy used and incorporating them in the review process. Using advanced multilevel designs will still only partly solve the methodological issues involved in studying environmental attributes associated with physical activity, but it will help in disentangling this complex relationship. Therefore, it is recommended that, whenever there is a presumably relevant grouping (context; eg neighbourhoods) in a study, a multilevel approach should at least be considered.
Statistics from Altmetric.com
Competing interests: None declared.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.