Avoiding Blunders When Analyzing Correlated Data, Clustered Data, or Repeated Measures
Authors' conclusion
It is important to model the additional correlation that occurs in correlated data.
Study context
- Population
- Patients with rheumatoid arthritis (RA) between 1988 and 2007 (subset of the 2017 study from Raheel et al)
- Participants
- 633 patients
- Comparison
- Generalized linear models (GLM) treating observations as independent vs mixed-effects models accounting for correlation
Methodological detailssecondary analysis of a subset of the 2017 study from Raheel et al, GLM vs generalized linear mixed model with random intercept and generalized estimating equation
- Secondary analysis of a subset of the 2017 study from Raheel et al
- GLM vs generalized linear mixed model with random intercept and generalized estimating equation
Primary result
As reported in the abstract
β coefficients and SEs, RA flare and number of swollen joints
- β coefficients compared between methodologies are very similar
- SEs increase when correlation is accounted for
- If additional correlations are not considered, the SE can be underestimated: overestimated effect size, narrower confidence intervals (CIs), increased type I error and smaller P value
Original publication
Avoiding Blunders When Analyzing Correlated Data, Clustered Data, or Repeated Measures. J Rheumatol. Published 15 May 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →