リッカート型項目データの回帰への使用と通常最小2乗推定量

概要

論文の詳細を見る
Likert-type surveys are widely used in the field of social science, and data of individual Likert-type items, called Likert-type item data in this paper, are used for regression analysis as well. Doubt is put on the reliability of such regression analysis. It is shown that Likert-type item data contain measurement errors, and the use of such data for explanatory variables in regression makes the ordinary least squares (OLS) estimator biased and inconsistent. This is the same as the problem of "errors in variables" in regression literature. The problem gets worse if Likert-type item data under consideration do not have qualification of interval data and such qualification is not checked by researchers. To deal with the problem of "errors in variables," the popular estimation method is the instrument variable method. However, it is impossible to find instrument variables in the case of Likert-type surveys. A Likert-type survey is conducted since such variables are not available. One method for dealing with this situation is the use of dummy variables. In other words, it is the digitalization of data obtained from a Likert-type survey. Such digital data do not contain quantitative measurement errors. Unless the classical assumptions are violated, the OLS estimator is the best unbiased linear estimator.
青森公立大学の論文
2004-03-20