Open
Description
A measure ("test") added to the input with data appears in the output as an acceptable candidate but is not scored.
We use the detected measures in the input for generating candidates, but the list of loaded measures form the KB for scoring. This is a potential source of anomalous behavior.
- If we are limiting acceptable, scored measures to what is in the KB then we should mark other measures in the input as invalid.
- If we are handling any measures in the input (regardless of whether they exist in the KB measures collection) we should not filter
Right now We have updated esteemer to use the same list of measures as candidate_pudding (those for which motivating information was generated based on the input performance data). We need to apply the business rule above if it is needed.