Ego sleep mediates relationship between alter sleep and ego drug use.

Table 3 shows mediation effect of sleep on drug use. First three columns show logistic regression model of ego's marijuana use behavior (the outcome variable) on alter's sleep behavior (the explanatory variable). Second three columns show logistic regression model of ego's sleep behavior (the mediator variable) on alter's sleep behavior (the explanatory variable). Last three columns show logistic regression model of ego's marijuana use behavior (the outcome variable) on ego's sleep behavior (the mediator variable) controlling for alter's sleep behavior (the explanatory variable). Models were estimated using a general estimating equation with clustering on the ego and an independent working covariance structure. Models with an exchangeable correlation structure yielded poorer fit. Fit statistics show sum of squared deviance between predicted and observed values for the model and a null model with no covariates. A bootstrap procedure that takes into account uncertainty of both the effect of the independent variable on the mediator and the mediator on the outcome variable shows that when an alter starts sleeping ≤7 hours, it increases the likelihood of ego drug use by 4% (95% C.I. 1% to 7%) via its effect on the mediator (ego sleep), which represents approximately 20% of the total effect of alter sleep on ego drug use.