Browsing by Author "OLUWAFEMI, Alexander Olayinka"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Open Access
    Effectiveness of missing data methods in detecting differential item functioning in achievement motivation measurement among secondary school students in Osun state, Nigeria.
    (Department of Educational Foundations And Counselling, Faculty Of Education, Obafemi Awolowo University., 2021) OLUWAFEMI, Alexander Olayinka
    The study investigated the extent to which missing data methods (Full Information Maximum Likelihood (FIML), Multiple Imputation (MI), Simple Regression Substitution (SRS), and Relative Mean Substitution (RMS) were able to detect differential item functioning in achievement motivation measurement. It determined the significance of differential item functioning magnitude across missing data methods. It also ascertained the type of differential item functioning detected across missing data methods and estimated the type 1 error rate of differential item functioning detection across missing data methods. Furthermore, the study established the statistical power of Item Response Theory-Likelihood Ratio Test (IRT-LRT) method of differential item functioning detection across missing data methods. These were with a view to determining the occurrence of differential item functioning in polytomously scored non-cognitive tests. The study adopted descriptive Survey research design. The population for the study comprised all senior secondary school students in Osun State. The study sample consisted of 1,500 senior secondary school III students selected using multi stage sampling procedure. From each of the three Senatorial Districts, five Local Government Areas (LGAs) were selected using simple random sampling technique. From each of the selected LGAs, two schools which had not less than 50 SSSIII students each were purposively selected irrespective of school ownership. Every SSSIII students in the sampled schools were engaged to avoid the problem of keeping some students out of the classroom. An adapted version of the Achievement Motivation Inventory developed by Muthee and Thomas (2009) was used to collect data for the study. Data collected were analysed using percentages, mean, chi-square, ANOVA, Kruskal-Wallis statistics, MICE and MIRTincorporated in R programming language. The results showed that the four missing data methods considered were able to detect differential item functioning with Multiple Imputation method performing significantly better (F (3,124) = 34.48, p < 0.05) than the other three methods.The result further showed no significant different in the Differential Item Functioning (DIF )magnitude detected across missing data methods (χ2= 2.01, df = 3, p = 0.57). the result also showed that for most items (90.63 %), uniform DIF was detected with no interaction between the independent variables (school ownership, school location and sex) and group membership while multiple Imputation performed better in detecting non-uniform DIF across sub-groups. The result further revealed that no significantt difference among the DIF magnitude detected across missing data methods (χ2= 2.01, df = 3, p = 0.57). Likewise, the type one error rate did not differ significantly across missing data methods (ChiCrit (Bonf) = 11.35; ChiCrit (0.05) = 7.82) and there was no significant difference in the statistical power of DIF across missing data methods. The study concluded that there is occurrence of differential item functioning in polytomously scored non-cognitive test and missing data methods were effective in detecting differential item functioning in Likert-type non-cognitive scales. However, of the missing data methods considered, Multiple Imputation outperforms other methods in detecting DIF in non-cognitive scales.