- so far very small (10 items, 40 students)
- aim for 300 students next year
2018-10-19
Pilots
How are we going to evaluate this project?
- reliability
- validity
- usability
Reliability
- consistency of results
- are we measuring anyting?
Bayesian networks
updates of the student model
based on the task models
split half reliability
- if the items in the test measure the same thing
- there should be a relation between the results on two halfs of the test
Student model
Task models
Validity
- comparison with expert judgements
"replicate current practice" (Sangwin, this morning)
Validity
- diagnostic relevance
- what to report?