A Goodness-of-Fit Test for GEE Models with Binary Longitudinal Data Based on Smoothing Methods
DOI: 10.1109/ICICIC.2007.28 Conference: Innovative Computing, Information and Control, 2007. ICICIC '07. Second International Conference on
Source: IEEE Xplore
The logistic regression models have received widespread use for analyzing binary response data. In longitudinal studies, correlated data arise and such data are often analyzed by generalized estimating equations (GEE) method. This article proposes an alternative goodness-of-fit test based on nonparametric smoothing approach for assessing the adequacy of GEE fitted models, which can be regarded as an extension of the goodness-of-fit test of le Cessie and van Houwelingen (1991). The approximate expectation and variance of the proposed test statistic are derived. The power performance of test is discussed by simulation study and the testing procedure is illustrated by a clinical trial example.
- [Show abstract] [Hide abstract]
ABSTRACT: Most biomedical research is carried out using longitudinal studies. The method of generalized estimating equations (GEEs) introduced by Liang and Zeger [Longitudinal data analysis using generalized linear models, Biometrika 73 (1986), pp. 13–22] and Zeger and Liang [Longitudinal data analysis for discrete and continuous outcomes, Biometrics 42 (1986), pp. 121–130] has become a standard method for analyzing non-normal longitudinal data. Since then, a large variety of GEEs have been proposed. However, the model diagnostic problem has not been explored intensively. Oh et al. [Modeldiagnostic plots for repeated measures data using the generalized estimating equations approach, Comput. Statist. Data Anal. 53 (2008), pp. 222–232] proposed residual plots based on the quantile–quantile (Q–Q) plots of the χ2-distribution for repeated-measures data using the GEE methodology. They considered the Pearson, Anscombe and deviance residuals. In this work, we propose to extend this graphical diagnostic using a generalized residual. A simulation study is presented as well as two examples illustrating the proposed generalized Q–Q plots.Journal of Applied Statistics 11/2012; 39(11). DOI:10.1080/02664763.2012.710896 · 0.42 Impact Factor
Data provided are for informational purposes only. Although carefully collected, accuracy cannot be guaranteed. The impact factor represents a rough estimation of the journal's impact factor and does not reflect the actual current impact factor. Publisher conditions are provided by RoMEO. Differing provisions from the publisher's actual policy or licence agreement may be applicable.