Abstract
This chapter discusses how and why multilevel modelling is a flexible and powerful tool for analysing data with a hierarchical structure. Such data are often found in social science and public health research (e.g. when analysing pupils who are nested in classes or schools, patients in clinics or hospitals).
The aim of multilevel modelling is to integrate the regression equation at a lower level of data grouping (usually individuals) with that at higher levels (such as classes, schools, neighbourhoods) into one regression equation and to incorporate covariates at appropriate levels. By using multilevel models, it is possible to adjust for similarity of the lower level units belonging to the same group of a higher level and to make overall inferences about relationships between lower level as well as higher level characteristics and the outcome of interest.
Using example data sets, we explain step by step how to conduct linear or logistic multilevel modelling. Additionally, we provide syntax commands for several software packages and demonstrate how to interpret the results of multilevel analyses.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Note that every regression line is defined by only two points.
- 2.
Note: for the random effects in STATA and R, the standard deviation is given, while in SAS and SPSS, the variance (the square of the standard deviation) is given in the output.
References
Bates, D., Maechler, M., & Bolker, B. (2011). lme4: Linear mixed-effects models using S4 classes. 0.999375-42 ed. p. R package. Accessed February 11, 2013, from http://CRAN.R-project.org/package=lme4.
Bijleveld, C. C. J. H., & van der Kamp, L. J. T. (Eds.). (1998). Longitudinal data analysis. Designs, models and methods. London: Sage.
Diez-Roux, A. V. (2000). Multilevel analysis in public health research. Annual Review of Public Health, 21, 171–192.
Diez-Roux, A. V., & Aiello, A. E. (2005). Multilevel analysis of infectious diseases. The Journal of Infectious Diseases, 191(suppl 1), 25–33.
Dunson, D. (Ed.). (2008). Random effect and latent variable model selection (1st ed.). New York: Springer.
Freedman, D. A. (2004). The ecological fallacy. In M. Lewis-Beck, A. Bryman, & T. F. Liao (Eds.), Encyclopedia of social science research methods. London: Sage.
Graduate School of Education, Bristol Institute of Public Affairs. (2012). Centre for multilevel modelling. University-of-Bristol. AccessedFebruary 24, 2012, from http://www.bristol.ac.uk/cmm/.
Hox, J. (2002). Multilevel analysis. Techniques and applications. Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Hox, J., & Maas, C. J. M. (2004). The influence of violations of assumptions on multilevel parameter estimates and their standard errors. Computational Statistics & Data Analysis, 46, 427–440.
Inagami, S., Cohen, D. A., Brown, A. F., & Asch, S. M. (2009). Body mass index, neighborhood fast food and restaurant concentration and car ownership. Journal of Urban Health, 86(5), 683–695.
Kleinbaum, D. G., & Klein, M. (2010). Logistic regression. A self-learning text (3rd ed.). New York: Springer.
Koren, A., & Mawn, B. (2010). The context of unintended pregnancy among married women in the USA. Journal of Family Planning and Reproductive Health Care, 36(3), 150–158.
Kothari, A. R., & Birch, S. (2004). Multilevel health promotion research: conceptual and analytical considerations. The Canadian Journal of Nursing Research, 36(1), 56–75.
Kreft, I., & De Leeuw, J. (1998). Introducing multilevel modeling. London: Sage.
Larsen, K., & Merlo, J. (2005). Appropriate assessment of neighbourhood effects on individual health: Integrating random and fixed effects in multilevel logistic regression. American Journal of Epidemiology, 161, 81–88.
Lehmann, E. L. (1986). Testing statistical hypotheses (2nd ed.). New York: Wiley.
Matthews, S. A., & Yang, T.-C. (2010). Exploring the role of the built and social neighborhood environment in moderating stress and health. Annals of Behavioral Medicine, 39, 170–183.
Merlo, J., Chaix, B., Ohlsson, H., Beckman, A., Johnell, K., Hjerpe, P., et al. (2006). A brief conceptual tutorial of multilevel analysis in social epidemiology: Using measures of clustering in multilevel logistic regression to investigate contextual phenomena. Journal of Epidemiology and Community Health, 60, 290–297.
Montgomery, D. C., & Peck, E. A. (1992). Introduction to linear regression analysis. New York: Wiley.
Norman, C. D., Maley, O., Li, X., & Skinner, H. A. (2008). Using the internet to assist smoking prevention and cessation in schools: A randomized, controlled trial. Health Psychology, 27(6), 799–810.
Petree, R. D., Broome, K. M., & Bennett, J. B. (2012). Exploring and reducing stress in young restaurant workers: Results of a randomized field trial. American Journal of Health Promotion, 26(4), 217–224.
Pinheiro, J., Bates, D., DebRoy, S., & Sarkar, D., R-Development-Core-Team. (2011). nlme: Linear and nonlinear mixed effects models. R package version 3.1-102.
R Development Core Team. (2011). R: A language and environment for statistical computing. http://www.R-project.org, Vienna.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models. Applications and data analysis methods (2nd ed.). London: Sage.
SAS System for Windows. (2008). (9.2 ed.). Cary, NC: SAS Institute Inc.
Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis. An introduction to basic and advanced multilevel modeling. London: Sage.
SPSS for Windows. (2011). (20.0.0 ed.). Chicago: SPSS Inc.
StataCorp. (2007). Stata statistical software, Stata/IC (101st ed.). College Station, TX: StataCorp LP.
Stock, C., Bloomfield, K., Ejstrud, B., Vinther-Larsen, M., Meijer, M., Grønbæk, M., et al. (2011). Are characteristics of the school district associated with active transportation to school in Danish adolescents? European Journal of Public Health, 22(3), 398–404.
Torsheim, T., Currie, C., Boyce, W., Kalnins, I., Overpeck, M., & Haugland, S. (2004). Material deprivation and self-rated health: A multilevel study of adolescents from 22 European and North American countries. Social Science & Medicine, 59, 1–12.
Venables, W. N., & Ripley, B. D. (2002). Modern applied statistics with S. New York: Springer.
Verbeke, G., & Molenberghs, G. (2000). Linear mixed models for longitudinal data. New York: Springer.
Victor, R. G., Ravenell, J. E., Freeman, A., Bhat, D. G., Storm, J. S., Shafig, M., et al. (2009). A barber-based intervention for hypertension in African-American men: Design of a group randomized trial. American Heart Journal, 157(1), 30–36.
Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., & Smith, G. M. (Eds.). (2009). Mixed effects models and extensions in ecology with R. New York: Springer.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendix
Appendix
Syntax commands for the examples in STATA, R and SPSS. In R, we used the following libraries: nlme (Pinheiro et al. 2011), MASS (Venables and Ripley 2002) and lme4 (Bates et al. 2011).
Multilevel Linear Regression: The Intercept-Only Model
In STATA:
xtmixed perform || school:, var
In R:
library (nlme)
L1 <- lme (fixed=perform~1, random= ~1 | school, data=walk)
summary (L1)
In SPSS:
MIXED perform
/PRINT = SOLUTION TESTCOV
/FIXED = INTERCEPT
/RANDOM = INTERCEPT | SUBJECT (school) .
Multilevel Linear Regression: Random Intercept Model with One Individual-Level Covariate
In STATA:
xtmixed perform activ|| school:, var
In R:
library (nlme)
L2 <- lme (fixed=perform ~ 1+activ , random = ~1 | school, data=walk)summary (L2)
In SPSS:
MIXED perform WITH activ
/PRINT = SOLUTION TESTCOV
/FIXED = INTERCEPT activ
/RANDOM = INTERCEPT | SUBJECT (school).
Multilevel Linear Regression: Random Intercept and Slope Model with One Individual-Level Covariate
In STATA:
xtmixed perform activ|| school: activ, var
In R:
library (nlme)
L3 <- lme (fixed=perform ~ 1+active, random=~activ|school, data=walk)
Summary (L3)
In SPSS:
MIXED perform WITH activ
/PRINT = SOLUTION TESTCOV
/FIXED = INTERCEPT activ
/RANDOM = INTERCEPT activ| SUBJECT (school).
Multilevel Logistic Model for Binary Outcome: Intercept-Only Model
In STATA:
xtset school
xtlogit walkschool
In R:
library (MASS)
library (lme4)
L4<-glmer (walkschool~1+(1|school),family=binomial, data=walk)
summary (L5)
In SPSS:
GENLINMIXED
/DATA_STRUCTURE SUBJECTS=school
/FIELDS TARGET=walkschool
/TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=LOGIT
/FIXED USE_INTERCEPT=TRUE
/random use_intercept=true subjects=school.
Multilevel Logistic Model for Binary Outcome: Random Intercept Model with Individual-Level and School District-Level Covariates and Cross-Level Interaction
In STATA:
xtlogit walkschool perform less_farm perform_less_farm, or
In R:
library (MASS)
library (lme4)
L5<-glmer (walkschool~1+perform+less_farm+perform_less_farm+ (1|school),family=binomial, data=walk)
summary (L5)
In SPSS:
GENLINMIXED
/DATA_STRUCTURE SUBJECTS=school
/FIELDS TARGET=walkschool
/TARGET_OPTIONS reference=0 DISTRIBUTION=BINOMIAL LINK=LOGIT
/FIXED effects= perform less_farm perform_less_farm USE_INTERCEPT=TRUE
/random use_intercept=true subjects=school COVARIANCE_TYPE =UNSTRUCTURED.
Rights and permissions
Copyright information
© 2013 Springer Science+Business Media New York
About this chapter
Cite this chapter
Grittner, U., Bloomfield, K. (2013). Mathematical Approaches to Analysing Area-Level Effects on Health. In: Stock, C., Ellaway, A. (eds) Neighbourhood Structure and Health Promotion. Springer, Boston, MA. https://doi.org/10.1007/978-1-4614-6672-7_10
Download citation
DOI: https://doi.org/10.1007/978-1-4614-6672-7_10
Published:
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4614-6671-0
Online ISBN: 978-1-4614-6672-7
eBook Packages: MedicineMedicine (R0)