A second-order correction kept four ozone abundance mean signs in a 2024 invertebrate extract

We tested a second-order correction of ozone abundance log ratios in a public invertebrate extract. Four US and European publication-period samples had 69 records from 15 papers. Equal paper weights gave original means of +0.006, +0.242, +0.133 and -0.512. Corrected means were +0.023, +0.569, +0.120 and -0.519. All four signs stayed the same. One record with treatment relative standard error 2.204 drove 64% of the largest mean shift. A screen for imprecise means changed the earlier US sign and left only one earlier European paper. Only later Europe had an interval below zero under the primary weights; equal role weights made it include zero. Correction stability did not ensure stability under record selection. Small paper samples, broad outcomes, nonrandom reporting and uncertain experimental units limit inference. We test no climate interaction or causal population effect. Code, aggregate evidence, source hashes and independent checks are included.

ozone · invertebrates · observation bias · small samples · log response ratio · precision · regional replication