| Summary: | 碩士 === 逢甲大學 === 統計與精算所 === 97 === Randomized response technique (RRT) is commonly used to guaranty privacy and reduce the number of dishonest responses to sensitive questions in a survey research. The procedure yields a more accurate estimate of the proportion of the prevalent population. This article deals with logistic regression of data obtained from the unrelated question RRT (Greenberg et al. 1969) design when the covariate data are missing at random (MAR) or completely at random (MCAR). In particular, we compare the efficiency of the validation likelihood estimator using different estimates of the selection probabilities, which may be treated as nuisance parameters. Furthermore, we develop the large sample theory, and show that, they are more efficient than the estimator using the true selection probability. The proposed method is illustrated using data from a cable TV study in Taiwan.
|
|---|
