MCBS data would be analyzed using both descriptive and multivariate modeling. Given that outcome variables typically are categorical (no continuous outcomes), multivariate models would be estimated using dichotomous, multinomial, or ordered logistic regression equations. To account for the variance calculation complexities of the multistage MCBS sample design, SAS and SUDAAN software will need to be used. Survey weights will be applied specific to the survey rounds being used.
Cross-sectional differences could be examined in relation to the independent variables/factors listed above. Trends over time will be descriptive and include responses to questions that have been repeated for multiple years of the survey. Longitudinal analyses will focus on questions that have repeatedly been asked of the same respondent and can examine factors related to changes over time.