For a discussion of the impact of misspecification of the distribution of random effects on estimation of and inference about both the fixed effects and the random effects in multilevel logistic regression models see Austin [ 14].
The multilevel analysis tested five different models (see Table 3).
In order to test the derived hypotheses, multilevel models were specified (see, inter alia, de Leeuw and Meijer 2010; Hox 2010; Raudenbush and Bryk 2002; Snijders and Bosker 2012).
This is accomplished by specifying a multilevel latent class logit model (see e.g. McLachlan and Peel (2000) that yields the probability that, say, student i ((i = 1,2, ldots n)) in school g ((g= 1,2,ldots G)) in latent class c ((c = 1,2,ldots, C)) selected to receive the treatment.
The observed difference of 11 min in our sample is not significant, when applying the survey/panel method adjustment for multilevel modelling or the full-adjusted model (see figure 2).
All statistical analyses were conducted using R v2.92 [ 57], and models were developed using the glm and glmer functions for single-level and multilevel models respectively (for glmer see lme4 package [ 58]).
We estimate multilevel probit models of the probability of using endoscopic procedures and include person-level, PCSA, and county-level covariates, following our conceptual model (see Fig. 1).
Variables entered into models are listed in Section 3 and graphs of results show predictions derived from significant models; full estimates from these multilevel models are reported as supplementary data (see electronic supplementary material (ESM)).
Prepregnancy weight was obtained from the predicted weight at 0 weeks' gestation from multilevel models based on the antenatal weight measurements (see the following text).
