Fuzzy regression and traditional regression is applied to the simulation of estimating the latent trait

• Main Authors: Wan-Chun Yang, 楊婉君
• Other Authors: Yuan-Horng Lin
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/48378812323607554717

碩士 === 臺中師範學院 === 教育測驗統計研究所 === 91 === The paper is to compare the analysis of the latent trait by fuzzy regression and traditional regression. Since human thinking and behavior model exist uncertainty, the fuzzy linguistic questionnaire is proposed. This research estimate the data from fuzzy linguistic questionnaire and traditional questionnaire by fuzzy regression and traditional regression. It was supported that the latent trait is known, and the fuzzy linguistic variables is adopted to be triangular fuzzy number which is classified to assymmetry and symmetry. Two different methods of score calculating used are weight average of membership function for fuzzy linguistic variables and maximun of membership for traditional Likert Style Scale. Finally, using those two regression models estimate the results and compare the difference. The outcome shows that the number of people and questions is effective on the estimate of those two regression models. While the number of questions is less, the effect on the estimate of the fuzzy regression is better. On the other hand, the number of questions is more, the effect on the estimate of the triangular regression is better. However, the type of the fuzzy linguistic variables is not effective on the results. In the end of all, it provides some thinking on Computer-Adaptive Test and gives some suggestions to help the follow-research.