Number-Base Versus Methods Conversion Techniques for Teaching Number System Conversions in a Nigerian College of Education

ABSTRACT Number system conversions are a foundational topic in computer science education, yet the instructional organisation of the topic has received little empirical attention. This study compared two ways of organising instruction for Nigeria Certificate in Education (NCE) computer science students: the Number-base Conversion Technique (NCT), which organises teaching around source bases and the positional principles underlying each conversion, and the Methods Conversion Technique (MCT), which organises teaching around three general procedures (repetitive, summation, and substitution) applied across bases. The complete raw mark records of eight NCE I cohorts in the course CSC 114 (Number System) at the Federal College of Education (Technical), Gusau were analysed: four cohorts taught with NCT (2016/17 to 2019/20, n = 679) and four taught with MCT (2021/22 to 2024/25, n = 452), N = 1,131; the 2020/21 session was lost to the COVID-19 disruption. The population, all NCE I computer science students who sat the course in the eight sessions, was enumerated in full, so no sampling was required. Outcome data were the official, internally moderated examination records of the College, and data extraction was validated against each mark sheet’s own grade-summary totals. The outcome was the end-of-course total mark (0-100). No difference between techniques was found at any level of analysis: pooled student-level, t(1129) = 1.21, p = .23, Cohen's d = 0.07 (NCT M = 52.22, SD = 22.73; MCT M = 50.57, SD = 21.84); rank-based, Mann-Whitney z = -0.20, p = .84; and session-level, with the cohort as the unit of analysis, t(6) = 0.52, p = .62. A sensitivity analysis excluding the highest-scoring cohort (2016/17) slightly reversed the direction of the pooled difference (d = -0.05, p = .44), confirming its fragility. A significant association between technique and letter-grade distribution, chi-square (5) = 21.18, p < .001, Cramer's V = 0.14, is attributable to session-varying grading and moderation practices rather than to performance, given the null score results. The study concludes that the organisation of instruction, by base or by method, makes no detectable difference to examination performance in this setting, and documents how two earlier partial analyses of subsets of these data produced spurious advantages in opposite directions. Hybrid instruction and a randomised comparison with retention and transfer measures are recommended. Keywords: Number System Conversions; Computer Science Education; Teaching Techniques; Quasi-Experimental Design; Cohort Effects.