What a Term-Test Item Analysis Is Telling You
You teach Grade 8 science in a semi-urban school and set a 25-item single-answer multiple-choice term test, one mark each, no negative marking, for two classes totalling 80 students. The summary statistics are: mean 17.4, standard deviation 3.8, KR-20 reliability 0.61. Item 7 has a facility of 0.86, but the point-biserial is close to zero and slightly negative; inspection shows that the highest-scoring students chose distractor C. Item 19 has a facility of 0.31, and 28 of the 33 students in the bottom third chose distractor B while 14 of the top 20 chose the key. Two scripts from the same class show identical patterns of wrong answers on nine consecutive items. A parent whose daughter scored 14 out of 25 writes demanding a re-mark, arguing the paper was too hard and that school-based assessment marks should be added so that she passes. The principal asks whether the test can be used for class ranking and for selecting the zonal science quiz team.
1. Interpret the statistics. What do items 19 and 7 and the reliability figure each tell you, and what action follows?
Read the worked answer
Read the three figures together, not separately.
- KR-20 = 0.61 means the test is only moderately reliable. With 25 items and SD 3.8 on 25 marks, differences of one or two marks between students near the middle are essentially noise. Fine for whole-class diagnosis; weak for individual high-stakes ranking.
- Item 19 (facility 0.31) is hard but appears to be working: the key attracted 14 of the top 20 while 28 of the low group were pulled to distractor B. That pattern suggests B is a plausible, teachable misconception - it is diagnostic gold, not a fault. Check the wording for ambiguity, then use B in reteaching.
- Item 7 (facility 0.86, discrimination near zero/negative) is defective. When high scorers pick C, either C is defensible or the key is wrong. Read the item and the textbook source before deciding.
Action: re-key item 7 if C is defensible, otherwise delete the item and rescore all 80 scripts on 24 marks; recompute mean, SD and reliability. Never simply add marks to the complaining student - the correction must apply to everyone. Report the corrected distribution to the principal and record the rationale in the assessment file.
2. How do you answer the parent's request, and how should school-based assessment marks relate to this test mark?
Read the worked answer
Answer the request in three parts, in a meeting rather than by letter.
- Re-marking. Agree to a re-check, and define it: the whole cohort's scripts are checked against the key, and any key error is corrected for all. Offer to show the parent the script and the item analysis for the items her daughter missed. This is fair and also protects you from a charge of making an exception.
- 'Too hard'. Show the distribution: mean 17.4, and her daughter at 14 - below average but within the main body of the class, not an outlier. The test was not unreasonable; it was, however, only moderately reliable, so near the cut score treat the mark with caution.
- SBA as a top-up. Explain plainly that SBA marks are not a bonus added to reach a pass. SBA is a separate, criterion-referenced record of specified practical work and assignments, combined with term-test marks only in the ratio set by the school's assessment policy and the relevant circular. If her daughter's SBA record is incomplete, the remedy is to complete the specified tasks - not to invent or inflate marks. That distinction protects the child's future results as much as the school's integrity.
Close by agreeing the next step in writing: what she will do, what the school will do, and by when.
3. What is the correct process for the two identical scripts?
Read the worked answer
Nine identical wrong answers in a row is unlikely by chance, but a pattern is not a finding. Treat it as an unresolved allegation.
Process: keep the scripts safely; compare against the seating plan; look for other shared patterns; separately ask each student to explain two of the shared wrong answers orally, with a third person present, and record their words. If explanations are absent or identical in error, refer the matter to the principal under the school's examination-misconduct procedure. Notify both sets of parents in writing, give the students a right of reply, and decide only after the reply.
What you must not do: announce a zero, shame the students in class, or drop the matter because it is inconvenient before the term report. Two failure modes are equally bad - a rushed penalty that cannot be defended, and silence that teaches the class that copying is safe.
Whatever the outcome, change the routine: alternate versions, spaced seating, and supervision that moves rather than sits.