Copyright (c) 2009 All rights reserved. http://works.bepress.com/justine_shults Recent documents in Justine Shults en-us Sun, 31 May 2009 07:58:07 PDT 3600 http://works.bepress.com/justine_shults/3 http://works.bepress.com/justine_shults/3 Thu, 26 Oct 2006 11:28:58 PDT In a recent publication, Wang and Carey (Journal of the American Statistical Association, 99, pp. 845-853, 2004) presented a new approach for estimation of the correlation parameters in the framework of generalized estimating equations (GEE). They considered correlated continuous, binary and count data with a generalized Markov correlation structure that includes the first-order autoregressive AR(1) and Markov structures as special cases. They made detailed comparisons with pseudo-likelihood (PL) and the first stage of quasi-least squares (QLS), a two-stage approach in the framework of generalized estimating equations (GEE). In this note we extend their comparisons for the second (bias corrected) stage of QLS. We comment on their earlier findings, which were overwhelmingly in favor of the Wang-Carey (WC) approach relative to stage one of QLS. We prove that WC and QLS are identical for equally spaced data with an AR(1) structure. Furthermore, we demonstrate via simulations that neither QLS, PL or WC is uniformly superior for unequally spaced data with a Markov structure. We give general recommendations regarding the relative merits of each approach for analysis of unbalanced and unequally spaced longitudinal data and demonstrate their application in an analysis of a longitudinal study of obesity following renal transplantation in children. Wenguang Sun Longitudinal Data Analysis and Time Series http://works.bepress.com/justine_shults/1 http://works.bepress.com/justine_shults/1 Thu, 26 Oct 2006 11:28:57 PDT Justine Shults General Biostatistics http://works.bepress.com/justine_shults/2 http://works.bepress.com/justine_shults/2 Thu, 26 Oct 2006 11:28:57 PDT It is well-known that the correlation among binary outcomes is constrained by the marginal means, yet approaches such as generalized estimating equations (GEE) do not check that the constraints for the correlations are satisfied. We explore this issue for Markovian dependence in the context of a GEE analysis of a clinical trial that compares Venlafaxine with Lithium in the prevention of major depressive episode. We obtain simplified expressions for the constraints for the logistic model and the equicorrelated and first-order autoregressive correlation structures. We then obtain the limiting values of the GEE and quasi-least squares (QLS) estimates of the correlation parameter when the working structure has been misspecified and prove that misidentification can lead to a severe violation of bounds. As a result, we suggest that violation of bounds can provide additional evidence in ruling out application of a particular working correlation structure. For a structure that is otherwise plausible and results in only a minor violation, we propose an iterative algorithm that yields an estimate that satifies the constraints. We compare our algorithm with two other approaches for estimation of the correlation that have been proposed to avoid a violation of bounds and demonstrate that it estimates the correlation parameter and bivariate probabilities with smaller mean square error and bias, especially when the correlation is large. Justine Shults Longitudinal Data Analysis and Time Series