Statistical Methods
Audit a Lower-Triangular Linear Solve
Define the task
A triangular solve appears inside many regression and covariance algorithms. Auditing it can reveal an incorrect matrix orientation or a wrongly assumed unit diagonal. This tool solves a small real lower-triangular system Lx=b using its supplied diagonal. It does not invert a matrix, fit a regression or determine whether an upstream factorization was appropriate. Preserve row order and the numerical scale of both coefficients and right-hand side.
Carry out the check
Starting with the first row, subtract contributions from already solved components, then divide by the current diagonal coefficient. Repeat downward. The tool rejects nonzero entries above the diagonal, zero diagonal entries and nonfinite results. It reports the maximum absolute residual in Lx−b as an arithmetic check. A small residual alone does not prove a stable solution: an ill-conditioned system can greatly amplify small input changes. Do not silently replace a small diagonal with one.
Worked hypothetical example
For fictional L with rows (2,0) and (3,4), and b=(2,11), the first component is 1 and the second is (11−3)/4=2. Multiplication reconstructs (2,11) exactly in this simple example. The teaching tool accepts the matrix rows followed by one right-hand-side row. For a real analysis, retain the factorization method, precision and conditioning assessment alongside the solution. Transpose and upper-triangular conventions require a different operation and are not inferred automatically.
Method checked 19 November 2026 against Primary method documentation: solve_triangular.