Effectiveness of Classical and Generalizability Test Theory in Estimating Reliability of Teachers Job Performance Scale in Secondary School in Oyo State

Abstract

The study estimated the reliability of the teacher job performance scale using classical test theory. It also examined the reliability of the teacher job performance scale using generalizability theory. Furthermore, it compared the margin of measurement error in the teacher job performance scale based on classical and generalizability test theories. These were with a view to improving the psychometric quality of the teacher job performance scale.  Descriptive survey research design was adopted for the study. The population for the study consisted of all secondary school teachers in Oyo State. The study sample consisted of 450 secondary school teachers who were selected using multistage sampling technique. Three Local Government Areas (LGAs) were selected from each of the three senatorial districts in the state using simple random sampling technique. Five secondary schools were selected from each LGA using simple random sampling technique. Furthermore, 10 teachers were selected in each school using simple random sampling technique. The research instrument titled teacher job performance scale (TJPS) was used to collect the data. The TJPS was adopted from Goodman and Svyantek (1999) job performance scale developed by Goodman and Svyantek (1999). The reliability of the TJPS was estimated using Cronbach alpha, Spearman Brown Split halve and Pearson test retest for CTT and G and Phi coefficient for GT. The results showed that the reliability of the teacher job performance scale based on classical test theory was 0.86 (Cronbach’s alpha), 0.80 (split halve) and 0.685 (test retest). On the other hand, the reliability of teacher job performance scale based on generalizability theory was 0.86 (G-coefficient) and 0.85 (Phi-coefficient) for one-facet design, while the reliability was 0.625 and 0.59 for G and Phi coefficients respectively using two-facet design   Furthermore, estimated margin of measurement error from reliability coefficients based on classical test theory was 4.9% (Cronbach’s alpha), 5.8% (split halve) and 7.03%  (test retest), while the margin error based on GT was 0.18% and 0.29% using one and two-facet respectively. The study concluded that the reliability of teacher job performance scale can be improved using the generalizability theory.