Methodology
Every result comes from a rule and real computation.
Clarus is not a prediction engine but a rule-based analysis tool. Here are the six principles its reliability rests on.
- Rule-based analysis
Each method runs from a detailed definition; the result comes from a documented rule, not from memory. There is no language model in the analysis path.
- Choosing the right method
It weighs your design and your data and ranks the methods that fit your purpose, with reasons — you need not know which test to ask for.
- Assumption checks
Every assumption the chosen method requires is tested one by one. A violation is shown plainly, and a robust or alternative method is suggested.
- Verified computation
Calculations use scipy and statsmodels and are checked against known results; the accuracy of every number is verified.
- Source in view
Every important number traces back to the method, formula and input that produced it. The source sits one click away, without noise.
- Reproducibility
Every analysis comes with a publication-ready report and reproduction code; another person reaches the same result from the same data.