00000namaa22000004a 4500 UP-100084294495502615 Buklod 20230713150421.0 aa grm |||| u ta 230713s2022 ph a rm 00| 0|engdd DEDUC rda eng DMLUC LG 996 2022 E3 M66 Montemayor, Janet Lynn S. author Predictive framework for estimating the 3PL model parameters of mathematics test items Janet Lynn S. Montemayor ; dissertation adviser, Jose Q. Pedrajita ; dissertation reader, Monalisa Te Sasing ; Teodora M. Salubayba, Maricris B. Acido-Muega, Joselito C. Magadia, members. Quezon City College of Education, University of the Philippines Diliman 2022 xvii, 217 leaves illustrations 28 cm. text rdacontent unmediated rdamedia volume rdacarrier Dissertation (Doctor of Philosophy in Education -- Research and Evaluation) University of the Philippines Diliman February 2022 Includes bibliographical references (166-171) Accurately estimating the statistical parameters and the amount of information an item can provide is more essential during test construction to effectively fit the test with the examinees, thus ensuring that the instrument is more valid and reliable. There is no known established system that estimates the probable statistical attributes of test items. The primary objective of the study was to generate a predictive framework that provides statistical parameter estimates of test questions in Mathematics. The effect of item format, cognitive level, and context on the difficulty, discrimination, pseudo-chance level, and amount of information provided were analyzed using Item Response Theory. Other physical attributes that influence the statistical parameters of test questions were determined using content analysis. Mixed methods research design, particularly the embedded experimental model, was employed. A 48-item Mathematics test was administered to 2 231 Grade 7 students from seven public high schools. The three- parameter logistic (3PL) model was utilized to determine the statistical parameters of each item. Six teachers and 30 students were interviewed to validate the physical attributes identified. Mathematics questions in multiple-choice format were less difficult, less discriminating, and more prone to guessing than short-answer items. Cognitive level and context did not significantly affect the statistical parameters of items. The amount of information obtained about examinees was greater for short-answer items, items requiring higher order thinking skills, and those written in the local context. These classifications of test items provided more information about and are therefore most suitable for examinees with ability levels of average or above. Physical attributes that influence statistical parameters of test items include the form of stimuli (i.e., mathematical expression, illustration, and solution) and type of options. From the results of quantitative and qualitative data analyses, decision trees that provide estimates of difficulty, discrimination, and pseudo-chance level indices are generated. These decision trees are imperative in the field of Assessment and Evaluation as it is a potential material in appraising the statistical parameters of items during test development. Mathematics Examinations, questions, etc. Mathematics Philippines Baguio City Examinations. Mathematics Philippines La Trinidad (Benguet) Examinations. Mathematics Philippines Testing. Mathematical ability Testing. Item response theory. Pedrajita, Jose Q. dissertation adviser. Sasing, Monalisa Te dissertation reader/critic. Salubayba, Teodora M. member. Muega, Maricris B. Acido member. Magadia, Joselito C. member. FI UP UPD DEDUC LG 996 2022 E3 M67 Thesis