A bank with features | Predict difficulty, with honest uncertainty | How many pretest responses do we need? | Calibrate from a small pretest | Which families can we trust? | Items of a family share prediction error

Data | Calibrate each group, then link and detect DIF together | Does DIF change pass rates? | How much do the conclusions depend on the linking? | Two contrasting cases | Reading the verdict | What should translators look at? | A report to start from

A simulated administration | Fit the many-facet Rasch model | The decision rule is the point | Counterfactual pass probabilities | Where do wrong decisions come from? | Checking against the truth | Oral examinations: one examiner per case, scores 0-100

Two cohorts | Link interims to the summative scale | Routing policies | Fairness | Can a through-year score replace the summative?

Data | Window estimates | Set the alarm threshold for your bank | Monitor | Act, and record why | What does drift do to scores?

Data | Expected gain | Evidence and risk | Compared with raw-gain flagging