Experimental units, randomisation, blocking, mixed models, effect sizes, multiplicity and machine-readable data publication.
Experimental unit
The biological experimental unit is the smallest independently assigned unit. Multiple leaves from one plant are subsamples, not independent plants. Plants sharing one reservoir may not be independent for a nutrient-solution treatment; reservoir or run can become the true treatment unit.
Design controls
- random assignment to position and treatment;
- blocking by cycle and relevant spatial factor;
- blinded sample coding for laboratory and sensory analysis;
- predefined exclusion and deviation rules;
- power analysis updated from pilot variance;
- effect size and confidence interval reported with p-values.
Example mixed model
ExplanationThe model separates treatment, position, cultivar and cycle effects so a treatment is not credited for variation produced elsewhere.
Multiple outcomes
Primary outcomes are declared before analysis. Secondary chemistry and sensory endpoints are labelled and multiplicity is controlled where inferential claims are made.
Measurement uncertainty and precision terminology are declared separately from biological variation. Repeatability and reproducibility are not treated as synonyms.
Data package
Every report links raw data, processed data, analysis code, data dictionary, protocol, deviations and checksums. Metadata follow MIAPPE concepts and FAIR principles.