estimation of measurement error in binary data in surveys

Measurement error is an inherent and unavoidable component of nonsampling error in surveys‎, ‎and its estimation is essential for assessing the quality of survey results‎. ‎this paper investigates two agreement criteria‎, ‎cohen's kappa and gwet's ac1‎, ‎for quantifying measurement error in binary survey data‎. ‎a theoretical comparison is first presented to highlight the conceptual differences between the two coefficients‎. ‎this is followed by a simulation study in which true binary values are generated under varying prevalence levels‎, ‎and observed responses are obtained through controlled misclassification mechanisms characterized by specified levels of observed agreement‎. ‎the performance and empirical variances of both agreement measures are evaluated and the results demonstrate that gwet's ac1 provides more stable agreement estimates and variance behavior than cohen's kappa‎, ‎particularly under conditions of extreme prevalence.