CONSISTENCY OF POSTGRADUATE ACADEMIC PERFORMANCE AT MUET, JAMSHORO: A WELCH EQUIVALENCE-TESTING AND EXPLORATORY REGRESSION STUDY

Abstract

Universities need evidence on whether the academic performance of successive postgraduate batches is stable and whether programmes differ, yet non-significant test results are often misread as proof of consistency. This study examines the CGPA records of 206 students at Mehran University of Engineering & Technology (MUET), Jamshoro: 53 in M.Phil Applied Mathematics (five batches) and 153 in MS/MBA at the Mehran University Institute of Science, Technology and Development (MUISTD; four batches), for the batches of 2014–2017. Batch means were compared with Welch’s t-tests (Holm-adjusted), Welch’s one-way ANOVA, Hedges’ g with 95% confidence intervals, and two one-sided equivalence tests (TOST) with a ±0.30 CGPA margin; a sensitivity analysis excluded zero-CGPA records. Mean CGPA ranged from 3.45 to 3.72 (Applied Mathematics) and from 3.37 to 3.59 (MUISTD). No comparison between batches (Holm-adjusted p ≥ 0.95 in Applied Mathematics and 1.00 in MUISTD) or between programmes (≥ 0.42) was significant, and the omnibus tests agreed (p = 0.42 and 0.61). However, equivalence within ±0.30 CGPA was demonstrated for only 4 of 20 comparisons (16 of 20 within ±0.50), so “no significant difference” cannot be equated with “consistency” at these sample sizes. Excluding zero-CGPA records raised the MUISTD batch means to 3.50–3.69 and brought the omnibus test close to significance (p = 0.064). A batch-level regression of on-time graduates on enrolment and CGPA fitted well in sample (R2 = 0.95) but was dominated by two high-leverage batches (Cook’s D = 12.9 and 18.3) and failed leave-one-out validation, so it is reported as exploratory. Student-level data are needed for confirmatory inference.