An assessment of the utility of a Bayesian framework to improve response propensity modes in longitudinal data