A Study on Extreme Range of Item Difficulty Parameters for Item Selection in Mokken Scale Analysis

• Main Authors: Gao,Chien Han, 高千涵
• Other Authors: 江振東
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/57385060040001089222

碩士 === 國立政治大學 === 統計研究所 === 101 === In a questionnaire survey, if the items are related to politics or personal privacy, respondents tend to give a negative response to the items. On the other hand, if the items are highly recognized, most of them answer positively to the items. These two are examples that the responses to the items are too consistent. It is apparent that useful information is hard to come by through the use of such kind of survey data. The automated item selection procedure (AISP) is a tool to select items that form a Mokken scale from questionnaire. However, the items selected by AISP is not necessarily the optimal choice. Therefore, the aim in the study was to investigate whether extreme response items would create more problems when using AISP. Number of items which violates the selection rules was based to judge the possible impact. The simulation results showed that number of violations increased when the items were too consistent. Furthermore, items with low discrimination almost led to ineffective items selection.