In this study we used a multilevel model (especially a three-level model) suggested by several researchers, including DerSimonian and Laird [72], [73], DerSimonian and Kacker [74], Goldstein, Yang, Omar, Turner, and Thompson [75], van den Noortgate and Onghena [76], Beretvas and Pastor [67], and van Houwelingen, Arends, and Stijnen [77].
Therefore, we used a multilevel model whereby ARCP outcomes were conceptualised as nested within individual doctors.
We used a multilevel model to assess variability in GDM between states.
To analyse consultation length, we used a multilevel model with three levels: contacts (1), GPs (2) and practices (3).
For the secondary analysis, including all four nights, we used a multilevel model, Prais regression, which accounts for the within-patient correlation between measurements on successive nights.
We used a multilevel model to estimate BMI at age 7 and the linear change in BMI between 7 and 13 years for all participants with one or more BMI measure.
We used a multilevel modeling (MLM) framework for estimation because we wanted to fit the regression to individuals while accounting for systematic, unexplained variation among counties and states.
We used a multilevel modeling strategy that we have described previously (Berhane et al. 2004).
Fourth, the rich data sample containing information at the patient and hospital levels allowed us to use a multilevel model combined with a 2SRI model.
Therefore, we use a multilevel model to examine these effects simultaneously.
We also tested whether there was significant practice variation using a multilevel model.
