Item Analysis
Item Analysis evaluates individual items in a test or questionnaire — difficulty, discrimination, and relationship to overall score. The core tool of classical test theory for refining assessments.
Step 1 — Provide your data
Matrix of respondents x items (numeric, e.g. 0/1 for correct/incorrect or partial credit scores).
Supports Excel Import, Sample Data Generator, and Manual Entry.
Step 2 — Results
Overall test statistics:
- Total score mean and SD
- Number of items and respondents
Per-item statistics:
- Item mean and SD
- Difficulty index (p) — proportion correct
- Discrimination index (D) — top 27% minus bottom 27%
- Point-biserial correlation — item vs total score
- Top-group and bottom-group performance
- Response distribution
- Automatic flags for problematic items
A UniversalChatBot is available for discussion.
Statistical methods used
Difficulty index
difficulty = item_mean / max_item_score
| Difficulty | Interpretation |
|---|---|
| < 0.20 | Too difficult (flagged) |
| 0.20 - 0.80 | Acceptable |
| > 0.80 | Too easy (flagged) |
Discrimination index (D)
D = top_27%_proportion - bottom_27%_proportion
| D | Interpretation |
|---|---|
| >= 0.40 | Excellent |
| 0.30 - 0.40 | Good |
| 0.20 - 0.30 | Acceptable |
| < 0.20 | Low (flagged) |
| Negative | Serious problem — miskeyed or confusing |
Point-biserial correlation — Pearson r between item score and total score. Below ~0.20 = weak item.
27% rule (Kelley, 1939) — top/bottom 27% maximises discrimination reliability.
No formal hypothesis test — item analysis is descriptive. Use alongside Cronbach's Alpha for comprehensive scale evaluation.