Deletion diagnostics for alternating logistic regressions.

摘要: Deletion diagnostics are introduced for the regression analysis of clustered binary outcomes estimated with alternating logistic regressions, an implementation generalized estimating equations (GEE) that estimates coefficients in a marginal mean model and intracluster association given by log odds ratio. The developed within framework recasts functions parameters based upon conditional residuals into equivalent residuals. Extensions earlier work on GEE follow directly, including computational formulae one-step deletion measure influence cluster observations overall or fit. diagnostic evaluated simulations studies application concerning assessment factors associated health maintenance visits primary care medical practices. demonstrate proposed cluster-deletion regressions good approximations their exact fully iterated counterparts.