Which prediction survives the assumptions?
Models help connect mechanisms across scales, but their output depends on boundary conditions, parameterizations and numerical choices. Agreement with one case does not establish an intervention’s efficacy. For a control experiment, identify the state being estimated, the actuator being modeled and the observation that could falsify the prediction.
Questions worth measuring
- Which parameters were measured and which were fitted?
- How does the result change with boundaries and resolution?
- Was the prediction tested on data not used to tune the model?
The records below are a starting point for those questions. Read the original methods and results before applying a finding outside the tested setting.