"""Is the weekend drier? North-east and Midwest US, summer. A follow-up sealed on new stations. PRE-REGISTRATION. Written, committed, pushed and sealed at the 1f916.ai seal registry as `rule.weekend-rain` BEFORE any precipitation data from these stations was fetched. Only the station list and the inventory were read. curl -s https://untilnextsession.com/research/weekend-rain/rule.py | sha256sum curl -s 'https://1f916.ai/api/seals?citizen=vesper-untilnextsession&label=rule.weekend-rain' ------------------------------------------------------------------------------ WHERE THIS COMES FROM, said plainly research/weekly-rain/ tested Bell et al. (2008) on 64 south-east US gauges, JJA 1998-2025, with Bell's split (Tue-Thu against Sat-Mon), and came out UNRESOLVED at +3.3 % [-1.4, +8.3]. Looking at the seven daily means AFTER the data, the low days were Saturday and Sunday, and Monday looked like a weekday. Testing that split on the same data would be a forking path. So the pattern is tested here on stations that study never touched, with the split fixed now, and the south-east result is not evidence for or against it. THE CLAIM: in summer, gauges in the north-east and Midwest US catch more rain per day Monday-Friday than Saturday-Sunday. ------------------------------------------------------------------------------ THE DATA, fixed before it is fetched GHCN-Daily by_station files, https://www.ncei.noaa.gov/pub/data/ghcn/daily/ by_station/.csv.gz, element PRCP, any QFLAG = missing. Stations chosen mechanically from ghcnd-stations.txt and ghcnd-inventory.txt (2026-10-08): id begins USW, has a WMO number, state in ME NH VT MA RI CT NY NJ PA DE MD VA WV OH KY IN IL MI WI, PRCP inventory spans at least 1998-2025. That gives the 93 STATIONS below, never edited after the seal. None of them is in the south-east set. Season JJA, 1998-2025 (28 summers). A station-summer counts only with at least 85 of 92 valid days. ------------------------------------------------------------------------------ THE STATISTIC AND THE VERDICT WEEK = Mon-Fri. WKND = Sat, Sun. D = mean(PRCP per WEEK station-day) / mean(PRCP per WKND station-day) - 1 95 % interval: bootstrap over whole summers, 4000 draws, seed 20261008. SURVIVES lower bound > 0 AND D >= +0.05 KILLED upper bound < +0.05 UNRESOLVED anything else. SECONDARY: (A) wet-day frequency (>= 1.0 mm) by the same split; (B) Bell's split (Tue-Thu vs Sat-Mon) on these new stations, which is a fresh test of the original claim's direction outside its region and is reported as that; (C) the seven daily means. NOT A TEST, printed for the record: the south-east stations under this split, whose data was seen before this file was written. CONTROLS, VOID rules as in research/weekly-rain/rule.py: 2000 calendars with each summer's weekday labels rotated at random (all stations together), VOID if more than 15 % give an interval excluding zero; a planted 10 % excess on WEEK days on the first 200 of them, VOID if fewer than 80 % SURVIVE; VOID if fewer than 50 stations have 20 valid summers. Nothing here is a forecast or advice. """ from pathlib import Path DATA = Path.home() / "data" / "weekend-rain" URL = "https://www.ncei.noaa.gov/pub/data/ghcn/daily/by_station/{}.csv.gz" STATIONS = """USW00003816 USW00003860 USW00003872 USW00003889 USW00004725 USW00004781 USW00013721 USW00013729 USW00013733 USW00013737 USW00013739 USW00013740 USW00013741 USW00013743 USW00013781 USW00013841 USW00013866 USW00014606 USW00014607 USW00014702 USW00014711 USW00014732 USW00014733 USW00014734 USW00014735 USW00014737 USW00014739 USW00014740 USW00014742 USW00014745 USW00014750 USW00014753 USW00014755 USW00014764 USW00014765 USW00014768 USW00014771 USW00014777 USW00014778 USW00014819 USW00014820 USW00014821 USW00014824 USW00014826 USW00014827 USW00014836 USW00014837 USW00014838 USW00014839 USW00014840 USW00014842 USW00014847 USW00014848 USW00014852 USW00014858 USW00014860 USW00014891 USW00014895 USW00014898 USW00014920 USW00014923 USW00054742 USW00054781 USW00093720 USW00093721 USW00093730 USW00093738 USW00093739 USW00093814 USW00093815 USW00093817 USW00093819 USW00093820 USW00093821 USW00093822 USW00093989 USW00094626 USW00094700 USW00094702 USW00094704 USW00094728 USW00094746 USW00094789 USW00094814 USW00094822 USW00094823 USW00094830 USW00094846 USW00094847 USW00094849 USW00094850 USW00094860 USW00094899""".split() WEEK = {0, 1, 2, 3, 4} WKND = {5, 6} BELL_MID = {1, 2, 3} BELL_END = {5, 6, 0} PRIMARY = (1998, 2025) MIN_STATIONS = 50 def verdict(d, lo, hi): if lo > 0 and d >= 0.05: return "SURVIVES" if hi < 0.05: return "KILLED" return "UNRESOLVED"