From February 2017, information about the work of the MRC/CSO Social and Public Health Sciences Unit, University of Glasgow is available and updated on the University of Glasgow website.

Multilevel Modelling for Public Health

Multilevel modelling is a recognised statistical technique for the analysis of hierarchical data. Despite this, it is still seen as an 'advanced' statistical method and not taught routinely alongside regression modelling. A number of software packages are now dedicated to multilevel modelling or have some capability. However, as its popularity grows there remain gaps in our knowledge regarding the theory and application of multilevel modelling. Together with Peter Groenewegen from Nivel we train researchers in the use of multilevel modelling through week long courses Health in Context: A course in Multilevel Modelling in Public Health and Health Service Research

We have developed models appropriate for the analysis of repeated classifications as part of our work on the Oslo Mortality Study. In our case this has been exemplified by individuals with area of residence observed at multiple time points. We have extended the multiple membership model to enable empirical estimation of the weights associated with each time point (as opposed to the prior specification of all weights necessary for the standard multiple membership model), but discuss the limitations of both multiple membership models and cross-classified models for such data. Consequently we introduce the correlated cross-classified model as a means of inducing a correlation between area effects at different times.
Other methodological contributions have included developing methodology for the application of multilevel event history (survival) models to large datasets including routine data, distinguishing between the inferences that may be drawn from fixed effect and random effect models, work on effective methods for modelling very small clusters and encyclopedia entries for hierarchical linear modelling and the intraclass correlation coefficient. Collaborative research has included applications to dentistry, periodontology, income inequalities and attitudes to redistribution, and a study of the relationship between ethnic density and food and physical activity environments (funded by the National Prevention Research Initiative NPRI).
Other activities include a review of multilevel modelling using the statistical software SPSS and developing training materials for multilevel modelling using MLwiN.



Hotchkiss J W, Davies C, Gray L, Bromley C, Capewell S, Leyland AH. Trends in cardiovascular disease biomarkers and their socioeconomic patterning among adults in the Scottish population 1995 to 2009: cross-sectional surveys. BMJ Open 2012;2:e000771

open access  


Hotchkiss J W, Davies C, Gray L, Bromley C, Capewell S, Leyland AH. Trends in adult cardiovascular disease risk factors and their socio-economic patterning in the Scottish population 1995-2008: cross-sectional surveys. BMJ Open 2011;1:e000176

open access  

Hotchkiss J W, Leyland AH. The relationship between body size and mortality in the linked Scottish Health Surveys: cross-sectional surveys with follow-up. International Journal of Obesity 2011;35:838-51

pubmed  open access  


Leyland AH. No quick fix: understanding the difference between fixed and random effect models [editorial]. Journal of Epidemiology & Community Health 2010;64:1027-8


Næss Ø, Leyland AH. Analysing the effect of area of residence over the life course in multilevel epidemiology. Scandinavian Journal of Public Health 2010;38:119-26

pubmed  open access  


Leyland AH, Næss Ø. The effect of area of residence over the life course on subsequent mortality. Journal of the Royal Statistical Society Series A 2009;172:555-78.

open access  


Leyland AH, Næss Ø. Multilevel modelling of the longitudinal influence of neighbourhoods on health [e-letter]. Journal of Epidemiology & Community Health Online 2007:18 January

open access  


Leyland AH. Assessing the impact of mobility on health: implications for life course epidemiology. Journal of Epidemiology & Community Health 2005;59:90-91



Leyland AH, Groenewegen PP. Multilevel modelling and public health policy. Scandinavian Journal of Public Health 2003;31:267-274



Keskimäki I, Karvonen S, Sund R, Leyland AH. Monitasomallien käyttö terveystutkimuksessa [Multilevel modelling in health research]. Sosiäälilaaketieteellinen Aikakauslehti [Journal of Social Medicine] 2001;38:327-335

Leyland AH, Goldstein H , (eds). Multilevel modelling of health statistics. Chichester: John Wiley & Sons, 2001.


Leyland AH, McLeod A. Mortality in England and Wales, 1979-1992, an introduction to Multilevel Modelling using MLwiN. MRC/CSO Social and Public Health Sciences Unit Occasional Paper no. 1, Glasgow, 2000

open access  

External Collaborators


  • cross-classified model A multilevel model in which units at one level lie within two non-nested classifications. An example would be individuals who live in neighbourhoods and attend hospitals, with there being no strict nesting of neighbourhoods within hospitals or of hospitals within neighbourhoods.
  • Multilevel modelling A form of regression analysis designed to estimate effects when data are clustered within units at higher levels e.g. survey respondents within households or areas, patients within hospitals etc.
  • multiple membership model A multilevel model in which units at one level may lie within more than one unit at a higher level. An example would be individuals receiving treatment from more than one doctor.
View all glossary entries