C.I. Enunwah, A.M. Akwa

Cross River State College of Education (NIGERIA)
This study investigated the development and standardization of achievement test in SS mathematics using item response theory (IRT) framework. The instrument captioned Multiple Choice Test in Algebraic Processes (MCTAPIM) was based on SS mathematics curriculum. One theme under algebraic processes became sufficient for the study because the IRT model used in the study demanded a uni-dimensional construct. Eleven research questions guided the study. A sample of 1557 students selected through proportionate stratified random sampling technique was used for the study. An initial pool of eighty seven (87) items was constructed, validated contently and trial testing, fifty (50) items survived the item analysis and were used for final administration employing the assistance of mathematics classroom teachers of each sampled schools. Both trial testing and the final administration were carried out with SS 3 students randomly drawn from secondary schools in Cross River state. One hundred (100) students (50) each from both public and private schools) were used for the trial testing while one thousand five hundred and fifty seven students were used for the final administration. The research questions were answered using descriptive statistics, while the hypotheses were tested at 5% level of significance using the lord’s Chi-Square and Students – t – test statistics. All these statistics were generated by the IRT soft ware WINSTEPS 3.72 version by Linacre (2011). The results from the study show 47 items from the 50 items of MCTAPIM made up the standardized version of MCTAPIM. Three items were dropped because their outfit mean square values were higher than 1.5 logit units. The instrument was suitable for IRT because IRT assumptions of unidimensionality and local independence of items were satisfied by the MCTAPIM data. Item difficulty parameters fall within – 0.55 and 3.13 logit units. The items separated the students into 15.84 strata according to item difficulty parameter. And Students’ responses were valid because 47 out of 50 items fit the model.