Statistical rigor,
within reach.
Most researchers have the data and the questions — but not a statistician down the hall. thericerca is built to close that gap: it chooses the correct test for what you’re asking, computes it for real, and explains what it means.
A deterministic engine maps your objectives to a statistical plan and selects the correct test, checking its assumptions as it goes. An optional AI can help read plain-English objectives — it never chooses a test or touches a number.
Every statistic is computed by vetted scientific libraries — scipy, statsmodels — in an isolated sandbox. Deterministic, inspectable, reproducible.
The gap thericerca closes is a quiet, expensive one. A well-designed study can still reach the wrong conclusion because the wrong test was run — a correlation where a chi-square belonged, an ANOVA on data that broke its own assumptions, a post-hoc comparison that was never licensed by its omnibus result.
General-purpose AI made this worse, not better. Ask a chatbot to “analyze my data” and it will happily return numbers that look authoritative and are quietly fabricated. For research, a plausible number is worse than no number at all.
How thericerca is built
At its center is a deterministic statistical engine — a rules engine that selects the appropriate test from your variables and objectives, and a registry of vetted analysis functions that actually run it. The engine checks each test’s assumptions and falls back when they fail: non-normal data moves Pearson to Spearman, unequal variances move ANOVA to Kruskal–Wallis. Post-hoc comparisons appear only when the overall test earns them.
A language model is used, but kept on a short leash. It helps translate your objectives into a plan and rewrites the findings into readable prose. It is never allowed to produce a statistic. Every number in your report is traced back to the computed results by a numeric-claim validator before the report is issued — if a figure can’t be verified, the report is flagged rather than shipped. thericerca runs its deterministic path with no AI at all; the model only enriches the wording when AI assistance is enabled.
What you can defend
The result is a report you can put in front of an advisor, a reviewer, or a client: the test chosen and the reason it was chosen, the assumptions checked, the results in plain language beside their formal statement, and the exact Python that produced every figure. You see the whole plan and its cost before you spend anything.
thericerca is an instrument, not an oracle. It automates the mechanics of choosing and running the right test with fidelity — it does not replace the judgment of a qualified statistician or domain expert, and it isn’t a substitute for professional, legal, or clinical advice.