Orthogonal Second-order Locally Ancillary Estimating Function

• Activity Number: 406
• Type: Contributed
• Date/Time: Thursday, August 15, 2002 : 10:30 AM to 12:20 PM
• Sponsor: General Methodology
• Author(s): Molin Wang*+ and John Hanfelt
• Affiliation(s): Emory University and Emory University
• Keywords: Bias correction ; Nuisance parameter ; Estimating function ; Errors-in-covariate ; Locally ancillarity

Abstract

We consider a general estimating function that involves not only the parameters of interest, but also some nuisance parameters. We provide an orthogonality condition under which a second-order locally ancillary estimating function can reduce its plug-in bias, caused by substituting an estimator for unknown nuisance parameter, to the order of o (1). According to this condition, we propose a class of second-order locally ancillary estimating functions for matched pair studies where the stratum-specific intercepts are treated as nuisance parameters, and second-order locally ancillary estimating function for errors-in-covariates problems where the mismeasured unknown covariates are treated as fixed nuisance parameters. The proposed estimating function for errors-in-covariates is applicable when the information about the second moment of the data is incomplete. The class of proposed estimating functions for matched pair studies involve neither the dispersion parameter nor information about higher than second moment of the data. Theories and simulation studies show that the proposed estimating functions are satisfactorily insensitive to nuisance parameters.