Changes of five percent of the series maximum allowed monotone fits in 27 of 34 ozone dose series in a 2024 data extract

We tested changes in direction in matched ozone dose series from public invertebrate data. Strict matching kept 34 series from 13 publications and 91 elevated-treatment rows in a 2024 release. The supplied means changed direction in 22 series. An exact error bound showed that changes to each mean of up to 5% of its series maximum permit a monotone fit in 27 series. Seven series need a larger change. All 34 permit a monotone fit within conservative sampling intervals under the stated mean and sample-size assumptions. In the same 22 series with at least three elevated doses, 14 change direction with the control and seven without it. Separate region and publication-period groups have uneven support; the later Americas group has one publication. These are fixed-data sensitivity results, not measured error rates, proof of biological monotonicity, or evidence that ozone is safe. Selected outcomes, uncertain experimental units, shared records and missing exposure duration limit inference. No climate or temperature effect is tested. The package gives original code, aggregate evidence, a short proof, source hashes, a figure and independent numerical checks. Source records and derived treatment tables remain external.

ozone · invertebrates · dose response · monotonicity · observation bias · sensitivity analysis · sampling uncertainty · regional replication