Live public record · August 18, 2026
Applied Statistics
An open, executable methods library connecting estimands, derivations, software parity, diagnostics, identification boundaries, and defensible communication.
- Status
- Live public record
- Topics
- statistics · econometrics · open education · reproducibility
- Boundary
- The linked artifact documents the current public release. Research findings require their own paper or dataset record.
Version 0.1 is one complete vertical slice rather than an empty curriculum shell. Its two-group inference module provides purpose-based Learn it / Run it / Defend it / Teach it routes; compares pooled, Welch, and binary OLS with HC2; and mechanically verifies the exact coefficient and standard-error equivalences.
The module keeps estimation separate from identification. Its browser lab defaults to descriptive language for observational assignment and makes causal wording conditional on the design that would support it.
The evidence base includes a deterministic synthetic fixture and an archived, attributed World Bank public-data case comparing 2024 life expectancy across low-income and upper-middle-income economies. JavaScript, Python, R, and Julia implementations are executed in the release gate. Excel and Stata versions are not claimed because they are not yet included and mechanically tested.
Question
How much can one observation change a fitted line and its uncertainty?
Add, move, or remove observations. The fitted line, confidence band, slope, robust standard error, and R² update together.