A comparison of bias-adjusted generalized estimating equations for sparse binary data in small-sample longitudinal studies.