Item analysis
Question statistics
| Question | n | Difficulty p | Discrimination r_pb | Verdict |
|---|---|---|---|---|
| What is the key difference between maximum likelihood (ML) and maximum a posteriori (MAP) estimation? r_pb=-0.49 < 0.15 — does not separate strong from weak | 6 | 0.83 | -0.49 | low discrimination |
| Why is a nonlinear activation function essential in a feedforward neural network with multiple layers? p=1.00 > 0.85 — almost everybody answers correctly | 6 | 1.00 | n/a | too easy |
| Using L2 weight decay (ridge regression) in a linear model corresponds to assuming which prior on the weights? | 6 | 0.67 | 0.61 | OK |
| What loss function does binary logistic regression minimize when using maximum likelihood? | 6 | 0.83 | 0.54 | OK |
| Why is gradient descent the standard method for training logistic regression models? | 6 | 0.33 | 0.16 | OK |
Rejection rule (from 5 attempts): a question is flagged when its difficulty p falls outside 0.30–0.85 (share answering correctly) or its point-biserial discrimination drops below 0.15 (correlation with the rest of the attempt score). Retiring and regenerating flagged questions runs offline: uv run python -m app.quality … --regenerate.