CHOICE TEST ITEMS

The purpose of this study was to examine the performance of five model-fitting estimation techniques; Likelihood Ratio Test (LRT), Akaike Information Criterion (AIC), Bayesian information criterion (BIC), Deviance Information Criterion (DIC) and Cross- Validation Log-Likelihood (CVLL) techniques that effectively selected an IRT dichotomous model which fitted NABTCE 2018 May/June Mathematics multiple choice test. This was carried out by comparing the performances of the five techniques used based on relative fit. Four research questions guided the study. No hypothesis was formulated and tested, due to the fact that the techniques used in this study were non-significant statistics. The research design employed was the descriptive survey of the ex-post facto method. The population of the study consisted of 49,581 candidates who sat for the National Business and Technical Certificate Examinations in 2018 in the six Geo-Political Zones in Nigeria. The sample size comprised 4,948 candidates and a statistical sample of 50 items. The Multistage simple random sampling technique was employed for randomly selecting the sample for the study. The instrument used to collect data was 50-item Mathematics multiple choice test from NBTCE May/June 2018. The instrument was a standardized instrument and as such it was valid and reliable. Item parameters were estimated from the examinees’ responses to the items using the computer programme BILOG-MG3. For the five estimation techniques BILOG-MG3 was used for LR, AIC and BIC. WinBUGS 1.4 was used for DIC, while MATLAB was used for CVLL which answered the research questions