Two sources in the corpus agree on the direction of a plant-microbiome result but disagree on how strong it is and whether its headline effect size can be used. The cross-project digest reports that co-occurring plant-associated genera show functional redundancy rather than complementarity, quoting permutation p = 1.0 and Cohen’s d = −7.54 (Cohen’s d is a standardized effect size: a mean difference divided by a standard deviation) 1. The plant-microbiome project later describes that same magnitude as largely a formula error 2. This matters for community-metabolic-interdependence. The finding contradicts the expectation that co-occurrence signals complementary metabolic provisioning, so the weight it carries depends on whether the number can be trusted 1.
Evidence Sides
Side 1: The digest reports redundancy with a very large effect. The digest states that co-occurring plant-associated genera show functional redundancy rather than complementarity (permutation p = 1.0, Cohen’s d = −7.54) 1. A permutation test compares an observed statistic with a null distribution generated by reshuffling the data. The quoted p-value does not itself support a significant permutation result, and that excerpt supplies no corrected estimate 1. As the digest records it, this side is a directional claim of redundancy with an extreme stated magnitude.
Side 2: The plant-microbiome project says the magnitude is an artifact and only the direction holds. The project describes the Phase 1 value from notebook NB06, cohens d = −7.54, as an implausibly large effect and largely a Cohen’s d formula error 2. NB06 divided the mean difference by the standard deviation of permutation means. By the central limit theorem, that quantity shrinks toward zero for any reasonable null. The correct denominator is the raw pair-level standard deviation, which is much larger 2. On this side, −7.54 should not be carried forward as an effect size, and only the direction survives 2.
Possible Reconciliations
- Hypothesis A: Both sources describe the same underlying computation. The digest captured the Phase 1 output before the correction, so the sources differ in timing rather than in data. Under this reading the redundancy direction is shared, and only the digest’s magnitude is outdated.
- Hypothesis B: The digest’s p = 1.0 reflects a test-tail convention rather than an absence of signal. The excerpt does not state which tail was tested.
- Hypothesis C: The redundancy direction holds, but the effect is too small to bear much weight against complementary provisioning. In that case the result neither strongly supports nor strongly contradicts interdependence at the community scale.
Resolving Work
- Replace the digest figures: Recompute Cohen’s d on the NB06 pair-level data, using the raw pair-level standard deviation as the denominator. Then update the digest entry so that it carries the corrected magnitude, or no magnitude at all, instead of −7.54.
- Check the test definition: Re-run the permutation test on the same co-occurring genus pairs and report the test tail and the null construction. This settles whether p = 1.0 describes a non-significant result or a reversed-tail artifact.
- Test aggregation sensitivity: Recompute complementarity under alternative species-level aggregation rules on the same NMDC (National Microbiome Data Collaborative) soil communities. This asks whether the redundancy direction is stable across methods.
- Test transfer to other communities: Apply the corrected effect-size formula to co-occurrence complementarity tests in other environments in the corpus. This asks whether any other reported co-occurrence effect sizes share the same denominator error.