GEE-smoothing spline for semiparametric estimation of longitudinal binary data

This paper considers analyzing longitudinal data semiparametrically and proposing GEE-Smoothing spline in the estimation of the parametric and nonparametric components. Generalized estimating equation is used as the core of the estimation. Estimation of association or within subject correlation used method of moment suggested by Liang and Zeger (1986). In the estimation of nonparametric component, we used smoothing spline approach specifically the natural cubic spline. We show through simulation that GEE-Smoothing Spline has good properties. The bias of parametric and nonparametric estimators decrease with increasing sample size. These estimators are also consistent even though incorrect correlation structure is used. The most efficient estimator can be obtained if the correct correlation structure is used rather than ignore the dependency.

Actions (login required)

View Item