Clarusstatistical analysis tool

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.

  1. 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.

  2. 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.

  3. 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.

  4. Verified computation

    Calculations use scipy and statsmodels and are checked against known results; the accuracy of every number is verified.

  5. 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.

  6. Reproducibility

    Every analysis comes with a publication-ready report and reproduction code; another person reaches the same result from the same data.