Why Raw Terminal Marks Mislead Ghanaian Families
An aggregate 8 from a lenient basic school does not carry the same latent ability as an aggregate 12 from a rigorous, strict mock examination. Here is how KyerɛMe normalises grading rigor to protect your child's CSSPS secondary school selection.
The Tale of Two 74% Marks in Integrated Science
Consider two candidates preparing for the BECE in Greater Accra: Kwame and Abena. Both score 74% on their second-term terminal reports in Integrated Science.
Kwame's School (Strict Private JHS)
The school tests using previous WAEC chief examiner marking schemes. The cohort average for Science is 48%, and only 3% of students scored above 70%.
Normalised Stanine: Grade 1 (Excellent)
Abena's School (Lenient Local Mock)
The teacher marks generously on internal homework and continuous assessments. The cohort average for Science is 76%, and over 55% of the class scored above 70%.
Normalised Stanine: Grade 4 / 5 (Average)
If Abena's parents use raw terminal scores to select Category A boarding schools (like Achimota or Presec), they face a severe risk of displacement on results day. KyerɛMe transforms raw percentages into standard percentile scores.
Our Four Methodological Pillars
1. Empirical Bayes Shrinkage (HLM)
In smaller basic schools with graduating classes of only 15 to 25 pupils, raw percentiles can be noisy and unstable. An exceptional student might dominate a weak class or vice versa.
We use Empirical Bayes hierarchical linear modeling to shrink school parameters toward regional and district category hyper-priors: \(\lambda = n / (n + \tau^2 / \sigma^2)\). As sample sizes grow, the school's own data dominates; for tiny cohorts, the model guards against overconfidence.
2. Ordered Probit Stanine Mapping
The West African Examinations Council (WAEC) does not grade BECE on percentages; it grades on a 9-point stanine scale (Grade 1 to 9), where Grade 1 represents the top 4% nationally.
Our engine uses ordered probit regression calibrated against national quota cut-points ('P(grade = k)'), yielding realistic distributions for best-6 aggregate scores (6–54).
3. Academic-year rules and school context
Admissions rules can change between academic years. Check the source, year and verification status alongside any rule used in an estimate.
Where a sourced rule distinguishes school types, the estimate needs to account for that context. Unverified rules must remain visibly provisional.
4. Commute & Day Catchment Guardrails
Selecting a Category D or Day option beyond a 15 km travel radius poses severe daily commute difficulties for candidates. Using our GhanaPost GPS and district centroid geocoder, we alert parents whenever a day selection exceeds feasible catchment distances.
What Statistical Modelling Cannot Do
We believe parents deserve full intellectual transparency. No predictive platform can replace the official examination council:
- Unanchored Mock Papers: If a school uses a custom private mock paper with unstandardised difficulty, individual term fluctuations may occur. Adding multiple terms of reports drastically improves predictive stability.
- Exam Day Variance: Illness, anxiety, or external exam-room conditions on the official WAEC BECE week introduce variance that historical classroom grades cannot capture.
- Statutory Dispositive Decision: Official placement is executed solely by the CSSPS Secretariat under the Ministry of Education. KyerɛMe provides advisory decision support; we do not issue placement letters or admission offers.