CurveMole website quick-start data

Synthetic Raman-like data: three area-normalised Gaussian bands, a quadratic
background, and independent Gaussian noise with known sigma_y = 2.
No experimental measurements or software-validation claims are implied.

raman-spectrum.csv: primary tutorial input; includes a one-sample +180 artifact
at 900 cm^-1. Exclude 894--906 cm^-1 before fitting.
raman-clean.csv: identical underlying signal and noise without the artifact.
raman-with-cosmic.csv: same values as the primary input, saved separately for
the optional cosmic-removal plugin exercise. Work on a separate imported copy.
Each input CSV has X, observed intensity and absolute one-standard-deviation
sigma_y columns. The original input files must remain unchanged.

raman-truth.csv: noiseless components, noise, original observations, the
three-point background estimate, and the fit-exclusion mask. This file is
ground truth for inspection, not the tutorial's fitting input.
reference.json: generation parameters, exact observed anchor coordinates,
mask interval and scientific conventions.

The quadratic background is estimated from three observed control points, then held fixed and subtracted. Fit and resampling uncertainties are conditional on this chosen background; its estimation uncertainty is not propagated into the band parameters.
