In many studies, categorical outcomes are measured repeatedly over time. Naturally, those outcomes are correlated and methods based on marginal models are popular choices to analyze such data. However, in reality, marginal models may not provide an appropriate estimation procedure due to lack of proper specification of joint models for outcome variables for repeated measures. As an alternative to marginal approaches, conditional models have also been developed in relatively few studies. However, conditional models are also inadequate to address the problem of modeling correlated data. Both marginal and conditional models fail to represent the underlying dependence in correlated outcome variables. In this study, a joint model is proposed in order to address the limitations of marginal and conditional models. A relative comparison of selected marginal approaches and the proposed model is examined for bivariate outcomes.