A Comparison of Treatment after Detecting insufficient effort respondings in Survey data through Simulation study

Abstract

Since insufficient effort responding(IER) in survey data caused by a variety factors threaten the validity of the data and lead to inaccurate research results, it is necessary to detect and treat IER before conducting the main data analysis is a necessary in order to produce more reliable research results. Nevertheless, it has been insufficient to discuss about treatments how to deal with the data after detecting IER. Accordingly, this purpose of this study is to provide methodological implications about more feasible treatments after detecting IER by comparing complete data analysis, casewise- deletion, and multiple imputation which are applicable after detecting IER through simulation in the context of analyzing categorical confirmatory factor analysis. The main results of this study are as follows. First, It is showed that the casewise deletion tends to perform better than other treatments in terms of goodness-of-fit index and the accuracy of parameter estimation. Second, complete data analysis tends to perform poorly in the goodness-of-fit index and showed inaccurate results in estimating the relationship between constructs. Third, It is showed that multiple imputation tends to perform better than analyzing complete data in the model fit index, and it is indicated to be relatively accurate in estimating the relationship between constructs. However, It was inaccurate in terms of estimating factor loadings when multiple imputation was applied. Based on the results, It has been discussed about more efficient treatments how to deal with the survey data after detecting IER.