State score signs did not reject four published ozone and warmth timing contrasts in US bird surveys in 1995-2023

We audited four published prior-minus-future ozone and warmth contrasts for recorded bird richness in US surveys in 1995-2023. A restricted state-score sign reference did not reject zero after correction for four tests. In the early West, the adjusted p value was 0.414, compared with 0.0029 under the published state-cluster t method. One state supplied 92.2 percent of squared restricted score among 13 states. The contrast estimate remained -0.946 species per 10 ppb per 2 C. The audit shows sensitivity to inference, not which method has the correct error rate. Sign symmetry, nuisance estimates, unequal state sizes and cross-state dependence limit the reference. Complete western sign enumeration removes simulation error only. Detection bias, monitor error and selected future records remain. The same records are reused; there is no independent biological replication or causal claim. Code, aggregate state scores, input hashes, a figure and numerical checks are included.

ozone · seasonal warmth · bird richness · observation bias · cluster inference · restricted score · sign reference · inference audit