Annual mean and within-plant lilac bloom trend estimates differed by -3.30 to +4.32 days per decade in four US samples from 1965 to 2008

We tested how the changing mix of reporting plants changed estimated first-bloom trends. Four US region and period samples used 10,103 lilac reports in 1965-2008. We compared equal-year annual mean trends with a common linear trend within the same plants. Their differences were +1.96, -3.30, +3.65, and +4.32 days per decade. The early western point estimates had opposite signs. The three available primary contrast intervals included zero after correction for four tests. The fourth variance estimate was negative, so its interval was unavailable. Additional conservative intervals included zero in all four samples. Long-record and common-plant restrictions changed some signs. An exact fitted decomposition assigns the contrast to the changing annual mean of fitted plant offsets. It does not identify a population trend or a causal climate effect. Unknown visits, selected plants, different lilac types and protocols, sparse later western years, and few uncertainty clusters limit inference. The package includes source hashes, selected events, monthly weather, code, all results, figures, and independent checks.

phenology · lilac · observation bias · sampling composition · time trend · fixed effects · climate indicators · reproducibility