Weighted estimating equation: modified GEE in longitudinal data analysis

Abstract

The method of generalized estimating equations (GEE) introduced by K. Y. Liang and S. L. Zeger has been widely used to analyze longitudinal data. Recently, this method has been criticized for a failure to protect against misspecification of working correlation models, which in some cases leads to loss of efficiency or infeasibility of solutions. In this paper, we present a new method named as ‘weighted estimating equations (WEE)’ for estimating the correlation parameters. The new estimates of correlation parameters are obtained as the solutions of these weighted estimating equations. For some commonly assumed correlation structures, we show that there exists a unique feasible solution to these weighted estimating equations regardless the correlation structure is correctly specified or not. The new feasible estimates of correlation parameters are consistent when the working correlation structure is correctly specified. Simulation results suggest that the new method works well in finite samples.

Keywords

Consistency / correlation / efficiency / generalized estimating equation (GEE) / longitudinal data / positive definite / repeated measures / weighted estimating equation (WEE)

References

[1] Albert P S, McShane L M. A generalized estimating equations approach for spatially correlated binary data: applications to the analysis of neuroimaging data. Biometrics, 1995, 51: 627−638

[2] Carroll R J, Ruppert D. Transformations and Weighting in Regression. New York: Chapman and Hall, 1998

[3] Chaganty N R. An alternative approach to the analysis of longitudinal data via generalized estimating equations. J Statist Plann Inference, 1997, 63: 39−54

[4] Chaganty N R, Shults J. On elimination the asymptotic bias in the quasi-least squares estimate of the correlation parameter. J Statist Plann Inference, 1999, 76: 127−144

[5] Crowder M. On the use of a working correlation matrix in using generalised linear models for repeated measurements. Biometrika, 1995, 82: 407−410

[6] Davidian M, Giltinan D M. Nonlinear Models for Repeated Measurement Data. London: Chapman and Hall, 1995

[7] Liang K Y, Zeger S L. Longitudinal data analysis using generalized linear models. Biometrika, 1986, 73: 13−22

[8] Liang K Y, Zeger S L, Qaqish B. Multivariate regression analysis for categorical data. J Roy Statist Soc Ser B, 1992, 54: 3−40

[9] Lunn A D, Davies S J. A note on generating correlated binary variables. Biometrika, 1998, 85: 487−490

[10] McCullagh P. Quasi-likelihood functions. Ann Statist, 1983, 11: 59−67 Modified GEE in longitudinal data analysis 353

[11] McCullagh P, Nelder J A. Generalized Linear Models. New York: Chapman and Hall, 1989

[12] Nelder J A, Wedderburn R W M. Generalized linear models. J Roy Statist Soc Ser A, 1972, 135: 370−384

[13] Núñez-Antón V, Woodworth G G. Analysis of longitudinal data with unequally spaced observations and time-dependent correlated errors. Biometrics, 1994, 50: 445−456

[14] O’Hara-Hines R. Comparison of two covariance structures in the analysis of clustered polytomous data using generalized estimating equations. Biometrics, 1998, 54: 312−316

[15] Rotnitzky A, Jewell N P. Hypothesis testing of regression parameters in semiparametric generalized linear models for cluster correlated data. Biometrika, 1990, 77: 485−497

[16] Segal M R, Neuhaus J M, James I R. Dependence estimation for marginal models of multivariate survival data. Life Data Analysis, 1997, 3: 251−268

[17] Shults J, Chaganty N R. Analysis of serially correlated data using quasi-least squares. Biometrics, 1998, 54: 1622−1630

[18] Shults J, Ratcliffe S J, Leonard M. Improved generalized estimating equation analysis via xtqls for quasi-least squares in Stata. Stata J, 2007, 7: 147−166

[19] Sun W G, Shults J, Leonard M. Use of unbiased estimating equations to estimate correlation in generalized estimating equation analysis of longitudinal trials. 2006, UPenn Biostatistics Working Papers. Working Paper 4.

[20] Sutradhar B, Das K. On the efficiency of regression estimators in generalized linear models for longitudinal data. Biometrika, 1999, 86: 459−465

[21] Wang Y G, Carey V J. Working correlation misspecification, estimation and covariate design: implications for generalized estimating equation performance. Biometrika, 2003, 90: 29−41

[22] Wang Y G, Carey V J. Unbiased estimating equations from working correlation models for irregularly timed repeated measures. J Amer Statist Assoc, 2004, 99: 845−852

[23] Wedderburn R M. Quasi-likelihood functions, generalized linear models, and the Gauss-Newton method. Biometrika, 1974, 61: 439−447

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg