The effects of correlation between class on parameter estimates in small group sizes of multilevel data
Multilevel analysis is usually performed on hierarchical data consisting of two or more levels. The highest level is the cluster level and the lowest level is individual level. The variation between levels is known as the intraclass correlation, ρ. This study discusses the effects of the intraclass...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universiti Kebangsaan Malaysia (UKM)
2013
|
| Subjects: | |
| Online Access: | http://irep.iium.edu.my/40161/ http://irep.iium.edu.my/40161/ http://irep.iium.edu.my/40161/1/JQMA_9%282%292013_paper.pdf |
| Summary: | Multilevel analysis is usually performed on hierarchical data consisting of two or more levels. The highest level is the cluster level and the lowest level is individual level. The variation between levels is known as the intraclass correlation, ρ. This study discusses the effects of the intraclass correlation on parameter estimation when the cluster size is small. Simulation studies were performed with 27 different conditions on group size, sample size and intraclass correlation value. Findings show that different ρ resulted in different effects on parameter relative bias. In conclusion, ρ with values greater than or equal to 0.2 and group size of more than 50 tend to give smaller bias to parameters estimates. |
|---|