Introduction to multilevel modelling | Ian Brunton-Smith
Download as PPT, PDF2 likes3,003 views
This document introduces multilevel modeling, explaining its purpose in accounting for complex data structures and how it examines individual and group relationships. It discusses various data structures, the benefits of a multilevel approach, and highlights the importance of context in modeling heterogeneity. Additionally, the document provides resources for further information on related topics.
Introduction to multilevel modelling | Ian Brunton-Smith
- 1. Session 1: Introducing multilevel modelling Multilevel Models
- 2. Outline • Introducing multilevel modelling – What is multilevel modelling – Data structures and multilevel questions – The benefits of a multilevel approach
- 3. What is Multilevel modelling? Models that explicitly account for the complex (grouped) structure of data •Model individual relationships, group relationships, and link between them •Adjust for increased similarities amongst observations from the same group (dependency) •Partition overall variation between individual and group Range of data structures can be accounted for... 3
- 4. Two-level nested structure 4 Area (level 2) Resident (level 1) Area 1 r1 Area 2 Area 3 r2 r3 r5r4 r8r7r6 r9 • Is there between area variability in residents fear of crime? • Does the effect of victimisation experience on fear vary across areas? Area Resident Fear of crime 1 1 High 1 2 Very high 1 3 Moderate 2 4 Very low 2 5 Low 3 6 Moderate 3 7 Very low 3 8 Very high 3 9 Low
- 5. Two-level Repeated Measures Structure 5 Pupil (level 2) Occasion (level 1) Person 1 t1 Person 2 Person 3 t2 t3 t2t1 t3t1 • Do students differ in their level of improvement in tests? • Can we explain these differences in progress with student characteristics? Person Occasion Test score Gender 1 1 30 Male 1 2 45 Male 1 3 53 Male 2 1 62 Female 2 2 80 Female 3 1 29 Male 3 3 55 Male
- 6. Person 1 t1 Person 2 Person 4 t2 t3 t2t1 t3t1 Person 6 t3t2 Person 3 t3t1 Person 5 t1 t2 t3 Class 1 Class 2 Class 3 Class 4 School 1 School 2 Area N-level Structure Schools (level 4) Pupils (level 2) Classes (level 3) Occasion (level 1) Area (level 5) Country (level 6,7...)
- 7. Other non-nested structures also possible 7 • Multiple membership • Spatial • Cross-classified Plus combinations...
- 8. Benefits of a Multilevel approach 1. Correctly account for complex data structures Where single level regression underestimates standard errors 1. Incorporate information on group level relationships Where aggregate analysis only examines group level, and individual analysis can ignore groups (or incorrectly treat group effects as individual effects) 1. Link context to the individual How individual relationships are moderated by broader context 4. Modelling heterogeneity Moving beyond typical focus on average relationship to also model variances 8
- 9. Useful websites for further information • www.understandingsociety.ac.uk (a ‘biosocial’ resource) • www.closer.ac.uk (UK longitudinal studies) • www.ukdataservice.ac.uk (access data) • www.metadac.ac.uk (genetics data) • www.ncrm.ac.uk (training and information)
- 10. Useful websites for further information • www.understandingsociety.ac.uk (a ‘biosocial’ resource) • www.closer.ac.uk (UK longitudinal studies) • www.ukdataservice.ac.uk (access data) • www.metadac.ac.uk (genetics data) • www.ncrm.ac.uk (training and information)