Directly strengthens the evidence boundary that analytical claims cannot exceed the calibration and validation domain.
The chapter distinguishes calibration, cross-validation, and external test-set validation and links model performance to sample coverage and reference-method accuracy.
A robust NIR calibration must include routine and unusual process states, span the relevant concentration range, use accurate reference measurements, and be tested against samples not used to fit the model; prediction error and outliers must be evaluated explicitly.
157
153
Chapter 12 — Building and validating NIR calibration models
Fresh September28 recheck: printed153–157/PDF159–163 supports calibration/validation boundaries. High correlation alone is not accuracy; average prediction error is not a guaranteed limit for every sample.
Capture evidence faithfully. Select one Source, record a usable locator, and preserve enough surrounding context to prevent misquotation.
Evidence
Paste the exact quote, paraphrase, statistic, definition, or source claim into Excerpt Text. Use Excerpt Type to distinguish quotation from interpretation.
Location and context
Record Locator Type, Locator Start, Locator End when needed, and Source Section. Explain the surrounding argument in Context.
Annotation
In Annotation, state why this evidence matters. Add approved Tags for retrieval and connect it to every relevant Literature Note.