Analyses are reported to identify appropriate strategies for modeling correlated outcomes in terms of available predictors, addressing what methods to use and what order to apply them. Analyses are conducted using extended linear mixed modeling (ELMM) of a continuous outcome treated as normally distributed, a count/rate outcome treated as Poisson distributed, a dichotomous outcome treated as Bernoulli distributed, a positive continuous outcome treated as exponentially distributed, and a trichotomous outcome treated as either multinomially, ordinally, or discretely numeric distributed. Whether non-constant or unit dispersions are appropriate is addressed as well as adaptive additive modeling, adaptive moderation modeling, and a comparison of results for extended variance modeling and direct variance modeling. The effect of considering random effects/coefficients on results generated using directly specified correlation structures is also addressed. A detailed specification of possible analysis strategies is provided as well as an assessment of ELMM for analyzing theory-based models and future work needed.