Book contents
- Frontmatter
- Contents
- Preface
- Acknowledgements
- 1 Introduction
- 2 Basic principles of multilevel analysis
- 3 What do we gain by applying multilevel analysis?
- 4 Multilevel analysis with different outcome variables
- 5 Multilevel modelling
- 6 Multilevel analysis in longitudinal studies
- 7 Multivariate multilevel analysis
- 8 Sample-size calculations in multilevel studies
- 9 Software for multilevel analysis
- References
- Index
7 - Multivariate multilevel analysis
Published online by Cambridge University Press: 26 March 2010
- Frontmatter
- Contents
- Preface
- Acknowledgements
- 1 Introduction
- 2 Basic principles of multilevel analysis
- 3 What do we gain by applying multilevel analysis?
- 4 Multilevel analysis with different outcome variables
- 5 Multilevel modelling
- 6 Multilevel analysis in longitudinal studies
- 7 Multivariate multilevel analysis
- 8 Sample-size calculations in multilevel studies
- 9 Software for multilevel analysis
- References
- Index
Summary
Introduction
A special feature of multilevel analysis is that it can be used to perform multivariate analysis. Multivariate analysis means that more than one outcome variable is analysed at one time. In the literature, multivariate analyses are often confused with multiple or multivariable regression analyses, in which the relationship between one outcome variable and more than one independent variable is analysed. That situation was discussed in Chapter 5. Multivariate analyses are not very common in medical science, but they are (for instance) widely used in psychology. Probably the most frequently applied multivariate technique is the multivariate analysis of variance (MANOVA), in which the average values of more than one continuous outcome variable are compared between groups. When a significant difference is found between groups, the next step is to examine which of the outcome variables differs between the groups, or, in other words, which of the outcome variables is related to the (group) determinant. When no significance difference is observed in multivariate analysis, this basically indicates that there is no significant relationship between the (group) determinant and the separate outcome variables as well. In this situation, the multivariate analysis can be seen as an efficient precursor of possible univariate analyses. When no multivariate relationship exists, univariate analysis does not necessarily have to be performed. When multivariate analyses is used in medical science, it is mostly used to analyse the relationship between one or more independent variables and a ‘cluster’ of outcome variables.
- Type
- Chapter
- Information
- Applied Multilevel AnalysisA Practical Guide for Medical Researchers, pp. 108 - 122Publisher: Cambridge University PressPrint publication year: 2006