Skip to content

Import Your First Dataset

Start with a small, representative FTIR, NIR, Raman, or UV-VIS dataset. Ten to fifty spectra are enough to verify file reading, axes, metadata, and plots before you import a full calibration set.

Use one of:

  • .csv with one row per sample and spectral variables as columns, or one explicitly confirmed unheaded coordinate/intensity spectrum
  • .jdx or .dx JCAMP-DX spectra
  • .npy or .npz arrays
  • .mat matrix data
  • One-dimensional Bruker OPUS, qualified Galactic SPC, qualified legacy Thermo OMNIC SPA/SPG/SRS, and qualified Renishaw WiRE WDF sources are accepted directly. If the file contains several results, choose the exact scientific result to bind. Thermo Paradigm/OMNICxi containers still require export to an admitted open format.

See Supported File Types for the full matrix, optional-algorithm version notes, HITRAN/HAPI extra, and currently unsupported vendor containers.

Import Checklist

  1. Open Data > Upload.
  2. Select the file or files.
  3. Check Files for the exact names and extensions received by SpectraSherpa.
  4. For CSV, review the recommended interpretation. Persist the explicit unheaded X/Y or decimal-comma profile when applicable.
  5. Check Metadata for inferred labels, units, and source information.
  6. Check Data Matrix for row count, column count, axis direction, and any explicitly preserved missing values.
  7. Save the dataset to My Dataset if you want to reuse it in workflows.

Calibration Data

For PLS or classification, make target values explicit. Prefer a CSV or metadata file with stable sample IDs rather than relying on row order alone.

Before training, confirm:

  • samples line up with reference values
  • replicate spectra are either grouped intentionally or separated intentionally
  • units are recorded
  • target columns are named clearly

Common First Workflow

Run a PCA template first. PCA quickly shows whether spectra cluster, whether preprocessing is needed, and whether any sample looks obviously misaligned or outlying.

After PCA looks reasonable, move to the workflow that matches your scientific question: PLS Calibration, Classification, SIMCA QC, or Peak Finding and Library Comparison.