Women tend to repeat reproductive outcomes, with past history of an adverse outcome being associated with an approximate twofold increase in subsequent risk. These observations support the need for statistical designs and analyses that address this clustering. Failure to do so may mask effects, result in inaccurate variance estimators, produce biased or inefficient estimates of exposure effects. We review and evaluate basic analytic approaches for analyzing reproductive outcomes, including ignoring reproductive history, treating it as a covariate, or avoiding the clustering problem by analyzing only one pregnancy per woman, and contrast these to more modern approaches such as generalized estimating equations with robust standard errors and mixed models with various correlation structures. We illustrate the issues by analyzing a sample from the Collaborative Perinatal Project dataset, demonstrating how the statistical model impacts summary statistics and inferences when assessing etiologic determinants of birth weight.
| Schisterman, Enrique F; Moysich, Kirsten B; England, Lucinda J et al. (2003) Estimation of the correlation coefficient using the Bayesian Approach and its applications for epidemiologic research. BMC Med Res Methodol 3:5 |
| Faraggi, David; Reiser, Benjamin; Schisterman, Enrique F (2003) ROC curve analysis for biomarkers based on pooled assessments. Stat Med 22:2515-27 |
| Schisterman, Enrique F (2002) Statistical analysis. Receiver operating characteristic (ROC) curve and lipid peroxidation. Methods Mol Biol 196:343-52 |
| Schisterman, Enrique F (2002) Statistical correction of the area under the ROC curve in the presence of random measurement error and applications to biomarkers of oxidative stress. Methods Mol Biol 186:313-7 |
