Debug random vs. conditional compliance #54
Description
In a previous experiment, we noticed that there were no differences in outcomes between scenarios that used random and conditional routine test compliance. This may have been caused by a bug in the experiment (eg, incorrect scenario generation). However, in case it's a bug in the model, we should do some testing to ensure that routine and conditional compliance are working as expected.
This could potentially take the form of unit tests, an experiment, or both. Any way, the goal is to check whether
- Random and conditional compliance generate different outcomes, all else being equal
- Simulated interventions (ie, comparing outcomes between scenarios with different screening rates) have different effects with random and conditional compliance
My intuition is that a given screening regime should be more effective with random compliance, because an agent with polyps/CRC who is noncompliant with one routine test could be compliant with a future one. However, this could depend on a lot of other things, like the screening rate. The bottom line is that random and conditional compliance should definitely result in different outcomes downstream, one way or another.