Bait-lamina test for assessment of polluted soils: Rough vs. Precise scales

[Display omitted] •The choice of scale for the bait-lamina test (BLT) influences statistical inference.•The five-point scale is a good surrogate for the bait consumption rate in the BLT.•A high percentage of partially consumed bait makes the two-point scale too coarse.•Partial bait consumption is twice as common in polluted as in unpolluted sites.•The three-point scale and ISO 18311 scale are less biased relative of the five-point scale. The bait-lamina test is one of the few available methods for measuring the functional activity of soil animals, which was standardized for soil health assessment by ISO 18311. The bait consumption is measured visually on discrete scales. A two-point scale (0 or 1) is used more often, but the threshold for bait consumption scores varies from study to study: 1 point is assigned either to a hole that has been perforated to any extent, or to a hole that has been at least half perforated, or to a fully empty hole. Less often, a three-point scale (0, 0.5, and 1) or a five-point scale (0, 0.25, 0.5, 0.75, and 1) can be used. We investigated how much the use of different scales can influence statistical inferences. We have established that a point on a five-point scale is directly proportional to the proportion of the mass consumption of bait, so in this scale the value of feeding activity is an acceptable surrogate for the bait consumption rate. We analyzed outcomes of the bait-lamina test for 7 measurement rounds at the control area and an area highly polluted with metals near the copper smelter. We assumed the most accurate five-point scale to be the reference and compared it to six less precise scales with fewer points (two three-point and four two-point scales). The difference between the reference and such rough scales depends on the percentage of intermediate points (0.25, 0.5, and 0.75). It turned out that intermediate values are by no means rare (from 11% to 100% of all non-zero values), and in the polluted area they are almost twice as common as in the control area (66% and 28%, respectively). Because of this, the relative difference between the reference scale and rough scales is 2–4 times greater at the polluted site than at the control site. This can lead to a bias in the effect size index: for the rough scales, the log Response Ratio can either double or decrease by one-third relative to the reference scale. This increases the probability of type I and II errors in statistical hypothesis testing. Thus, the outcomes o