Agriculture & Future Food Systems · Article
Review, Repeat and Choose the Next Change
Turn a growing record into a bounded comparison that distinguishes a promising observation from evidence for a useful change.
Preparation and related reading
Shared Foundations
Related articles
This article helps you
- Choose one justified cultivation question and a defined outcome.
- Distinguish independently treated growing units from repeated observations within one unit.
- Interpret a small numerical comparison without claiming an untested improvement.
Turn a problem into one question
A batch emerges unevenly. The grower considers changing the seed lot, depth, lighting and watering together. A better result might follow, but that bundle would not reveal which change helped. Start with the sowing, condition and emergence records, then choose a plausible limitation that can be examined safely and meaningfully.
State the question and outcome before starting another batch. For example, does a specified placement method improve emergence by a defined observation point under otherwise comparable care? Emergence is one outcome; later live condition, labour and useful production may reveal different trade-offs. The published symptom and emergence articles help define these measures.
Retain a comparison and record what changed
Compare the proposed practice with a defined reference practice, not just a memory of the previous batch. Keep seed identity, planting date, medium and other relevant care comparable where possible. Record any unavoidable differences. A previous season with different weather is informative history but may be a weak causal comparison.
Changing one factor helps isolate a question, although interactions can still matter. If two factors must change together, describe the comparison as that combined practice. Do not later attribute its result to only one component. More complex questions may require a designed factorial experiment and statistical help, which have their own methodological home.
Use independently treated growing units
A treatment applies to a growing unit, such as an independently managed tray or plot. Counting many seedlings inside one treated tray improves the description of that tray; it does not create many independent tray comparisons. This distinction prevents pseudoreplication, in which observations within one unit are mistakenly treated as repeated applications.
Where conditions vary with position, group comparable units into blocks and allocate the practices within those groups without choosing favourable locations for one practice. Random allocation helps avoid systematic placement bias. Keep units sufficiently separate for the intended comparison; shared watering or shading can make nominally separate units influence one another.
Read a small paired account
Imagine three hypothetical blocks, each containing two independently managed 20-seed trays. One tray uses the reference practice and one the revised practice, allocated within each block. At the same observation point, reference emergence counts are 12,15 and 13; revised counts are 15,16 and 15. The totals are 40/60 and 46/60.
The observed proportions are about 66.7% and 76.7%, a difference of 10 percentage points using the unrounded fractions. Within-block differences are 3,1 and 2 emerged seedlings. This simple summary describes the invented comparison. It supplies no statistical uncertainty, guarantees no future benefit and does not establish that later plant quality improves.
| Block | Reference emerged | Revised emerged | Difference |
|---|---|---|---|
| 1 | 12 | 15 | +3 |
| 2 | 15 | 16 | +1 |
| 3 | 13 | 15 | +2 |
| Totals | 40 of 60 | 46 of 60 | +6 of 60 |
Retain the repeated-unit results
Read the diagram as text
Three hypothetical paired blocks contribute to the totals without turning individual seedlings into independent tray replicates.
Parts
- Block 1: 12→15 of 20
- Reference and revised independently managed trays.
- Block 2: 15→16 of 20
- Same endpoint, with practice allocation within this block.
- Block 3: 13→15 of 20
- Retain this pair separately as well as in the total.
- Totals: 40→46 of 60
- Observed difference 6/60=10 percentage points.
- Check variability and quality
- No statistical certainty or future benefit is supplied.
Connections
- Block 1: 12→15 of 20 → Totals: 40→46 of 60: Add counts
- Block 2: 15→16 of 20 → Totals: 40→46 of 60: Add counts
- Block 3: 13→15 of 20 → Totals: 40→46 of 60: Add counts
- Totals: 40→46 of 60 → Check variability and quality: Interpret within the design
Look for limits and other outcomes
Review the distribution of results rather than reporting only the better total. Check missing observations, deviations, consistency of the endpoint and whether the intended change was actually applied. Appropriate statistical analysis depends on the design and measurement; a difference in sample proportions is not automatically a statistically supported treatment effect.
Also examine later survival, root and shoot condition, time spent, materials and resource use. A method could improve early appearance while increasing workload or damaging later growth. Retain an ambiguous or negative outcome: it may reveal a poor hypothesis, a weak measurement or another limitation rather than a failed growing effort.
Repeat before expanding the claim
Use the result to decide what to observe next. Repeating a comparable design in another batch or season tests whether the response persists under changed conditions. Preserve both the original and later records, including contrary evidence. Do not silently select only successful batches when describing a practice.
A local comparison supports a local decision at the strength of its evidence. Extension to another crop, medium, operating scale or climate needs relevant testing. Shared scientific-record and uncertainty Foundations supply the broader method; here the practical outcome is a justified next change and an honest account of what it achieved.
Worked examples
Compare the invented totals
What is the observed emergence difference in the hypothetical three-block account?
- Reference total:12+15+13=40 of 60 seeds.
- Revised total:15+16+15=46 of 60 seeds.
- Difference:(46/60−40/60)×100=10 percentage points.
The revised practice has 6 more observed emerged seedlings among 60 seeds, or 10 percentage points higher emergence in this invented account. Statistical confidence and later quality are not established.
Recognise the experimental unit
Are 20 seedlings measured within one tray 20 independent replicates of its watering practice?
- Identify where the practice was independently applied: the tray.
- Separate within-tray observations from independently treated trays.
- Plan repeated tray applications and comparable allocation if testing that practice.
No. The 20 plants describe one treated unit. They do not supply 20 independent tray-level comparisons.
Check your understanding
If depth and watering changed together, can a difference be attributed to depth alone?
Show answer and reasoning
No.
The tested contrast includes both changes and any interaction.
Is an observed 10-percentage-point difference automatically a statistically significant effect?
Show answer and reasoning
No.
Design, variability, sample size and appropriate analysis determine the strength of inference.
Should an unsuccessful repeat remain in the evidence record?
Show answer and reasoning
Yes.
Removing it would hide variation and overstate reliability.
Key ideas
- Define one useful question and outcome before changing the practice.
- Keep the comparison and actual implementation in the record.
- Distinguish independent growing units from observations within them.
- Repeat and examine trade-offs before extending a local claim.
References
Sources supporting the explanations and examples in this article.