Bibliographic citations
Cano, J., (2018). Análisis comparativo de la teoría clásica de los test y la teoría de respuesta al ítem aplicadas en evaluaciones informatizadas [Tesis, Universidad Nacional Agraria La Molina]. https://hdl.handle.net/20.500.12996/3094
Cano, J., Análisis comparativo de la teoría clásica de los test y la teoría de respuesta al ítem aplicadas en evaluaciones informatizadas [Tesis]. : Universidad Nacional Agraria La Molina; 2018. https://hdl.handle.net/20.500.12996/3094
@mastersthesis{renati/247106,
title = "Análisis comparativo de la teoría clásica de los test y la teoría de respuesta al ítem aplicadas en evaluaciones informatizadas",
author = "Cano Alva Trinidad, Jesús María",
publisher = "Universidad Nacional Agraria La Molina",
year = "2018"
}
Higher education institutions are increasingly implementing virtual evaluations via web in their various courses, in order to automate and measure more accurately the knowledge acquired by students. The objective of this research is to carry out a comparative study of the Classic Theory of the Test and the Theory of Response to the Item, which can be applied to the computerized tests with the purpose of evaluating and measuring the psychometric and statistical properties, focusing the reliability and validity of the evaluation tests. As a case study, we used a computerized test of 30 questions (items) that was applied to 775 students from a higher education institution enrolled in a Basic Statistics course in the 2016 II semester. The analysis of the reliability, resulted with a Cronbach's alpha of 0.8325 could indicate a good reliability for the computerized test, it was also corroborated with a Spearman-Brown correlation of 0.815. The analysis with the TCT, the difficulty index, identified three very easy questions (V7, V8 and V12) that were removed, while the discrimination index did not find any questions with problems. The assumption of the unidimensionality of the test using factor analysis was tested with an explained variance of the first factor of 24.7%. The binary logistic model of the TRI that best fit the data of the test was that of three parameters (3PL). The calibration process with the 3PL model allowed the withdrawal of questions V28 (discrimination index higher 0.65) and V8, V12, V16 and V18 (index of chance greater than 0.4). While all the questions were within the allowed range for difficulty index
This item is licensed under a Creative Commons License