Making Categories Interpretable in Regression: References, Contrasts and Interactions

Coding determines the model interpretation

Assigning 1, 2 and 3 can impose equal spacing and linearity. That assumption needs justification.

Practical framework

• Use k-1 indicators for nominal variables.

• Choose a meaningful reference.

• Compare ordinal trend with free categories.

• Interpret main terms conditionally with interactions.

Reporting

Report coding, reference, collapsed levels and contrasts.

Common mistakes

• Treating nominal codes as continuous.

• Changing reference after results.

• Collapsing sparse levels without rationale.

References

• UCLA categorical coding guide

• STROBE official checklist

Accessed 20 June 2026. General methodological information.