Based on the idea of #61, we want to conduct an integrative experiment across statement design points. Here we outline the experiment and a registration for it.
Preparation procedures
Sampling procedure
- Sample 10 starting design points at random.
- Sample 15 statements for each design point.
- Assign 100 participants to each design point using inverse weighting
- When a design point is finished, evaluate PQ common sense and train a model to predict other design points with all completed data.
- Compute model certainty on unsampled design points and select the next design point based on the one about which we are least certain.
- Register predictions for the new design point before starting to sample it.
Required features
We should finish these other things before launching
Based on the idea of #61, we want to conduct an integrative experiment across statement design points. Here we outline the experiment and a registration for it.
Preparation procedures
Sampling procedure
Required features
We should finish these other things before launching