Abstract
The study estimated the item parameters of the two forms of the non-equivalent anchor tests to be equated and examined the relative effectiveness of classical test theory equating methods in the non-equivalent anchor tests to be equated. The study adopted the Non-Equivalent Anchor Test (NEAT) design. The population of the study comprised 137083 Senior Secondary two (SS II) students in Osun State. Sample for the study consisted of 1080 students that were selected using a multistage sampling procedure. A total of 45 Senior Secondary 2 (SS II) students were selected from each school using the simple random technique. Two adapted instruments titled Mathematics Achievement Test Form A (MATFA) and Mathematics Achievement Test Form B (MATFAB) were used to collect data for the study. The instruments were adopted from 2014 and 2015 West African Examination Council Mathematics objective items which served as MATFA abd MATFAB respectively. The 2014 National Examination Council Mathematics objective test items were adopted. This served as the anchor items for both form A and B respectively. The reliability of the two instruments were established to be 0.79 for MATFA, and 0.75 MATFAB using Kuder-Richardson 20 (KR-20) formula. Data collected were analysed using IRTEQ software package, Common Item Program for Equating (CIPE) and R version (3.4.1) software. The results showed that the average difficulty and discrimination power of form A and form B Non-equivalent anchor test equated under CTT were (X = 0.53, X = 0.23) and (X =0.29, X = 0.27) respectively. The result further showed that Turker’s mean (TMEAN), Levine mean (LMEAN) and Braun-Holland Mean (BMEAN) were the most effective method of equating followed by Tucker’s Linear (TLIN) and Levine Linear (LLIN) method and finally by equipercentile method. The study concluded that mean equating methods were more effective compared to linear and equipercentile equating methods.