Location: Livestock and Range Research LaboratoryTitle: Mixing soil samples across experimental units ignores uncertainty and generates falsely precise estimates of soil biota effects on plants
Submitted to: New Phytologist
Publication Type: Other
Publication Acceptance Date: 9/1/2016
Publication Date: 1/1/2017
Citation: Rinella, M.J., Reinhart, K.O. 2017. Mixing soil samples across experimental units ignores uncertainty and generates falsely precise estimates of soil biota effects on plants. New Phytologist. 1-3.
Interpretive Summary: • Problem- A recent study used a simulation to identify the rate of statistical errors for two different methodologies for assaying soil biota effects on plant performance. They concluded neither method was reasonably capable of detecting differences between treatments. • Accomplishment- We identified technical and conceptual errors in their analysis and that the preferred methodology is robust to both forms of statistical errors while the other suffers from high rates of false positives.
Technical Abstract: A number of recent soil biota studies have deviated from the standard experimental approach of generating a distinct data value for each experimental unit (e.g. Yang et al., 2013; Gundale et al., 2014). Instead, these studies have mixed together soils from multiple experimental units (i.e. sites within a region or plots receiving the same treatment) and grown plants in greenhouse pots containing the mixtures to determine if soil biota effects on plants varied by region or treatment (e.g. Yang et al., 2013; Gundale et al., 2014). In a recent paper, we showed these mixed soil sample (MSS) bioassays are guaranteed to lead to underestimated P-values and incorrectly narrow confidence intervals, while, unsurprisingly, independent soil sample (ISS) bioassays that subject each plant to soil from a single experimental unit give correct inferences (Figure 2 of Reinhart & Rinella, 2016). These realities led us to conclude the MSS approach is “fatally flawed.”