Soil sample pooling generates no consistent inference bias: a meta‐analysis of 71 plant–soil feedback experiments
Warwick J. Allen, Sarah J. Sapsford, Ian A. Dickie
Abstract
There is current debate on how soil sample pooling affects the measurement of plant-soil feedbacks. Several studies have suggested that pooling soil samples among experimental units reduces variance and can bias estimates of plant-soil feedbacks. However, it is unclear whether pooling has resulted in systematic mismeasurement of plant-soil feedbacks in the literature. Using data from 71 experiments, we tested whether pairwise plant-soil feedback direction, magnitude and variance differed among soil pooling treatments. We also tested whether pooling has altered our understanding of abiotic and biotic drivers that influence pairwise plant-soil feedbacks. Pooling of soil samples among experimental units was used in 42% of examined experiments. Contrary to predictions, pooling did not affect mean pairwise plant-soil feedback effect size or within-experiment variance. Accounting for soil sample pooling also did not significantly alter our understanding of the drivers of pairwise plant-soil feedbacks. We conclude that there is no evidence that soil sample pooling systematically biases estimates of plant-soil feedback direction, magnitude, variance or drivers across many studies. Given the debate of whether to pool soil samples, researchers should be aware of potential criticisms and carefully consider how experimental design and soil pooling methods influence interpretation of experiments.