Individual-date flag removal had small effects on four US lilac spring slopes under fixed effects in 1965-2008

We tested whether removal of individual-plant date outlier flags changed the link between spring warmth and lilac first bloom. Four separate network and period fits used 9,622 records from the historic US archive in 1965-2008. We kept the same reporting units in each full and filtered fit, with unit and year effects and winter temperature control. Removal of 116 flagged records changed the four spring slopes by +0.052 to +0.155 days per C. All four approximate intervals corrected for four tests included zero and lay inside a chosen +/-0.5 day/C bound. This supports a small effect of this specific filter under the primary model. Some checks with evaluated records only, equal unit weights, or other flags did not meet that bound. Missing visits, unknown true event dates, coarse weather and few state clusters limit inference. The test does not show that flagged dates are wrong or that either slope is a causal climate response. The package preserves source hashes, selected records, weather, code, results, figures and reproduction instructions.

phenology · lilac · temperature · observation bias · quality control · outlier flags · reproducibility