A Comparison Model for Measuring Individual Agreement

• Activity Number: 66
• Type: Topic Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yuqing Tang*+ and Lawrence Lin and Samad Hedayat
• Companies: U.S. Food and Drug Administration and Baxter Healthcare Corporation and University of Illinois at Chicago
• Keywords: individual agreement ; mean squared deviation ; total-intra ratio ; Intra-intra ratio

Abstract

This paper proposes a general comparison model for assessing individual agreement of k = 2 raters evaluating n subjects with m = 2 replicated readings. Users can explore total-rater agreement relative to intra-rater agreement where any subset of the k raters can be selected in the numerator and denominator. Users are also allowed to compare intra-rater agreement among selected raters. Based on the ratio of mean squared deviations (MSDs), two comparative agreement indices, total-intra ratio (TIR) and intra-intra ratio (IIR) are proposed. The TIR is a non-inferiority assessment such that the differences of individual readings from different raters can not be inferior by a pre-specified margin to the differences of the replicated readings within raters. TIR can be used whether a reference exists or not. The method used by Food and Drug Administration(FDA) for evaluating individual bioequivalence under relative scale becomes the special case of our approach. The IIR is a classical assessment such that the precision of selected raters can be better, equal, or worse than that of other raters. The estimation and statistical inference of TIR and IIR are obtained through GEE method.