statisticslisted

Statistical analysis that states its n, its interval, and the assumption that would overturn its conclusion, using the Python standard library alone. Use when the question is whether a difference is real, how large a sample must be, whether a streak means anything, what a correlation actually supports, or whether an outlier is genuinely anomalous; when an analysis needs auditing for a p-value without an effect size, an in-sample result, or a best-of-many comparison reported as if one was tried; and for value at risk on returns, whether a model improvement is real and calibrated, whether a bet is positive expected value, and what threshold a recurring data check should fire at. Not for pipelines or warehouses (data-engineering), model building (machine-learning), or charts (dataviz).
anton-winter-arch/dotagents · ★ 1 · Data & Documents · score 72

Install: claude install-skill anton-winter-arch/dotagents

# statistics Statistical work that is correct, not merely computed, in whatever context the question arrives from. It answers a plain question cold - a spreadsheet someone sent, a run of wins, two variants that scored differently - and it is also the inference layer wherever "is this number real" gets asked. **It is organized around the errors, not around a menu of tests.** Computing a mean wrong is rare. A confident conclusion the analysis never earned is common: a p-value read as an effect size, a model scored on the data it was fit on, a backtest that quietly knew the future, a correlation reported as a cause. Every reference file here exists because one of those is easy to commit and hard to notice afterward. Floor is Python 3.10+ standard library, so nothing needs installing. `numpy`, `scipy.stats` and `pandas` are used when present and never required; their absence changes speed and the size of the test menu, not an answer. ## The method Four things travel with every result. An analysis missing any of them is not finished, and the auditor exists to say which one is missing. 1. **State `n`.** Every statistic reports the number of observations behind it, after exclusions. Rows dropped for being missing or non-numeric are counted and reported, never dropped quietly, because the drop rule is frequently correlated with the thing being measured. 2. **Quantify uncertainty with an interval.** A point estimate with no interval is a number, not a finding. Report