12 From benchmark to decision

R. Heller · Date: 2026-07-29 · Last changed: 2026-11-09

What happens when the data are real?

Chapter status: planned. The application below has not yet been carried out. No single-cell comparison or improvement over a default method is claimed.

  • Use the Chapter 10 grid to select a method and its hyperparameters.
  • Apply that choice to a single-cell dataset.
  • Compare against the field-default choice.
  • Report a null result honestly if the benchmark-guided choice does not win.
  • Closes with visualisation and honest-reporting standards: alt text, no colour-only encoding, seed disclosure, aspect-ratio discipline.
This book was built by the bookdown R package.