Text

theses

The significance of formative assessments has recently been underscored in the educational measurement literature. Formative assessments can provide more diagnostic information to improve teaching and learning strategies compared to summative assessments. Cognitive diagnosis models (CDMs) are psychometric models that have been developed to provide a more detailed evaluation of assessment data. CDMs aim to detect students' mastery and nonmastery of attributes in a particular content area. Another major research area in psychometrics is computerized adaptive testing (CAT). It has been developed as an alternative to paper-and-pencil tests, and widely used to deliver tests adaptively. Although the traditional CAT seems to satisfy the needs of the current testing market by providing summative scores, the use of CDMs in CAT can produce more diagnostic information with an efficient testing design. With a general aim to address needs in formative assessments, this dissertation aims to achieve three objectives: (1) to introduce two new item selection indices for cognitive diagnosis computerized adaptive testing (CD-CAT); (2) to control item exposure rates in CD-CAT; and (3) to propose an alternative CD-CAT administration procedure. Specifically, two new item selection indices are introduced for cognitive diagnosis. In addition, high item exposure rates that typically accompany efficient indices are controlled using two exposure control methods. Finally, a new CD-CAT procedure that involves item blocks is introduced. Using the new procedure, examinees would be able to review their responses within a block of items. The impact of different factors, namely, item quality, generating model, test termination rule, attribute distribution, sample size, and item pool size, on the estimation accuracy and exposure rates was investigated using three simulation studies. Moreover, item type usage in conjunction with the examinees' attribute vectors and generating models was also explored. The results showed that the new indices outperformed one of the most popular indices in CD-CAT, and also, they performed efficiently with the exposure control methods in terms of classification accuracy and item exposure. In addition, a new blocked-design CD-CAT procedure was promising for allowing item review and answer change during the test administration with a small loss in the classification accuracy.

ETD_6879

Ph.D.

Includes bibliographical references

by Mehmet Kaplan

doi:10.7282/T3J38VN0

ETD doctoral

The author owns the copyright to this work.