While the Crossrank score can indicate strength of the relationship in the analysis set, subject matter expertise is required to interpret the nature of the relationship.
Typically you want to find variables that are causal and controllable. Controllable variables should have a data group or these should be configured as “settings” to be isolated easily using the filters on the hyperlift page.
Here are some common types of relationships to consider in your review of tags:
- Causal but not Controllable, (ex., seasonal temperature affects ability to dry), these are “context variables” and should be decided in the analysis strategy
- Causal and Controllable, highest value, add to potential rule set
- Proxy, can be interesting as an indicator of the cause but without actually having the causal variable in Braincube
- Consequence, can be interesting for understanding the potential value of solving the problem.
- Coincidental, no value (i.e. time-related)