Compost Tea and the Confounded Tomato Plot
A garden club tests whether compost tea increases tomato yield. Plot A (8 plants) is watered with compost tea twice weekly; Plot B (8 plants) gets plain water. At harvest, Plot A yields 14.2 kg and Plot B yields 9.6 kg. Plot A is along the south-facing fence and receives about 2 more hours of direct sun; Plot B is partly shaded by a maple. Plot A also received extra water because the tea was applied with a hose. The club concludes compost tea caused a 48% yield increase and wants to sell the tea.
1. Is the 48% yield difference evidence that compost tea caused higher yield? List confounds and explain how each could inflate the result.
Read the worked answer
No. The comparison is confounded at the plot level.
The 48% figure is (14.2 - 9.6) / 9.6 = 47.9%, but it cannot be attributed to tea.
- Sunlight: Plot A gets about 2 more hours of direct sun. Light drives photosynthesis and fruit mass, so this alone could explain the gap.
- Water volume: Plot A received extra water. More water can increase fruit size and reduce blossom-end rot.
- Soil and drainage: Two different plots may differ in organic matter, pH, or drainage.
- Plant material and planting date: If the plots used different seedlings or dates, yield differs for reasons unrelated to tea.
- Pest and disease pressure: The shaded plot may have different moisture and disease levels.
- Pseudoreplication: There is one plot per treatment. The eight plants are subsamples, not independent replicates of the treatment. Plant-to-plant variation cannot estimate treatment error when the treatment is applied to a whole plot.
A defensible statement is that Plot A yielded more, but the cause is unidentified.
2. Design a controlled field trial. Include assignment, watering, and measurement details.
Read the worked answer
Design:
- Use multiple beds or plots (for example 8 pairs) and pair them by sun exposure, soil type, and drainage.
- Randomly assign one bed in each pair to compost tea and the other to plain water.
- Apply the same total volume of liquid at the same times; the only difference is whether the liquid contains compost tea.
- Use the same cultivar, planting date, spacing, and pest management.
- Apply tea to the soil, not the foliage, so leaf disease is not confounded.
- Measure fruit mass per plant and fruit number per plant over a defined harvest window.
- Weigh blind (the person weighing does not know the treatment) and record soil moisture, temperature, and visible pest damage.
- Analyze with a paired test or a mixed model that accounts for bed pairing.
Replicating across two seasons would test whether the effect is stable. The critical judgment is that the treatment must be randomized at the level at which it is applied-here, the bed-not just among plants within one bed.
3. After a better trial, tea plots yield 12% more, but 95% confidence intervals overlap. What can the club conclude, and what should they do before selling?
Read the worked answer
They cannot yet claim a reliable 12% benefit.
- Overlapping 95% confidence intervals mean the data are consistent with no effect, a small benefit, or even a small harm. The 12% estimate is too uncertain.
- A 12% difference may be practically meaningful, but statistical uncertainty must be reduced before marketing it.
- They should increase replication, run another season, and report the effect size with its confidence interval rather than a single percentage.
- Before selling, they need product consistency, microbial safety testing, labeling, and any required regulatory compliance. A garden trial does not establish a commercial product claim.
The scientifically honest conclusion is 'the trial suggests a possible benefit of about 12%, but the evidence is inconclusive.' Policy or sales decisions can still be made, but not on the claim that the effect is proven.