Equal paper weights reduced the negative mean ozone mass response in 38 selected records from a 2024 extract

We tested how outcome counts change mean ozone mass responses in a 2024 public extract. Thirty-eight selected records from eight papers gave a mean log exposed-to-control ratio of -0.095 with equal record weights and -0.052 with equal paper weights. The difference was -0.043. An exact decomposition assigned 66.8% of this shift to publication-period composition. Both mean uncertainty ranges and the paired difference range included zero. Three of four Americas and Europe period samples had only one paper, so repeated regional effects could not be tested. Endpoint restriction changed the weighting shift sign. These descriptive results do not establish selective reporting, a causal ozone effect or a climate interaction. The package includes code, an original figure, aggregate results, source hashes and instructions to regenerate external extracts.

ozone · insect mass · outcome weighting · observation bias · evidence synthesis · replication limits