SEM template regressing FOSQ, ESS, and ISI on treatment, after controlling for baseline scores and covariates. Rectangles at sides of figure represent scale items; circles represent latent variables; rectangles in the middle of figure are covariates. Square in the middle of the figure is the treatment variable. Label including/ending in “B” indicates baseline measure; Label including/ending in P indicates post-PAP measure. SEM = structural equation modeling; FOSQ = functional outcomes of sleep questionnaire (FOSQ-10); ESS = Epworth sleepiness scale; ISI = Insomnia severity index. COV = covariate. TREAT = treatment group (0 = traditional; 1 = DREAM)

Context in source publication

Context 1

... both outcomes, demographic variables and covariates that indicate differences following distribution to treatment groups will be entered as statistical controls. For the secondary outcome, FOSQ-10, ESS, and ISI scores will replace PAP adherence as the dependent variable, and Bayesian structural equation modeling (SEM) will be implemented to accommodate the latent structure of the questionnaires (i.e., questionnaire items specified onto unobserved factors; see Fig. 1 for model specification). Initially, confirmatory factor analyses (CFA) will be performed to select the best from a series of competing models (one-factor, correlated factor, and bifactor) to establish the optimal structure and relations of the questionnaires. Then paths from the established factor structure from the CFA will be regressed onto the treatment pathway, after controlling for demographic and other relevant covariates. ...