It is a common concern on error in measuring exposure variable in marketing, the social and behavioral sciences. It might due to inadequacies in measuring instruments or human error, so that the measures are often imperfect and contain measurement error. Thus, attenuation of effect and inconsistency will occur when the predictor variable is measured imprecisely. However, errors in variables can also arise even when the measurement process is accurate. An increasing acknowledgement of the importance of measurement error leads to a various investigation on alternative methods for correcting the measurement error in exposure. In this thesis, we mainly focus on the correction methods on logistic regression model. Correction methods based on maximum likelihood estimator with first and second...[ Read more ]

It is a common concern on error in measuring exposure variable in marketing, the social and behavioral sciences. It might due to inadequacies in measuring instruments or human error, so that the measures are often imperfect and contain measurement error. Thus, attenuation of effect and inconsistency will occur when the predictor variable is measured imprecisely. However, errors in variables can also arise even when the measurement process is accurate. An increasing acknowledgement of the importance of measurement error leads to a various investigation on alternative methods for correcting the measurement error in exposure. In this thesis, we mainly focus on the correction methods on logistic regression model. Correction methods based on maximum likelihood estimator with first and second order expansions are introduced. Besides, comparison between this method and the other methods including the one proposed by Rosner et al are also shown.