Claim at v1.2.0

split-temp-covid-growth

5a14c04bc8444f41df56b940a26d542422e9b54bd98675298597368ebd18c193

Data

{
  "has_outcome": true,
  "kind": "confirmation",
  "lineage": {
    "audit": "docs/lineage_audit.md; data/derived/lineage_audit.json",
    "class": "a",
    "class_meaning": "exposure or treatment only, not the outcome",
    "classes": [
      "a"
    ],
    "early_sources": [
      {
        "class": "a",
        "dataset": "GHCN-Daily v3 (NCEI), doi:10.7289/V5D21VHZ",
        "evidence": "Daily temperature (GHCN-Daily) is the exposure, not the outcome. It is weather, fetched for earlier heat claims and reused.",
        "fetched": "2026-11-12T20:23:55-05:00"
      }
    ]
  },
  "outcome": "split-temp-covid-growth",
  "prereg_branch": {
    "name": "prereg/split-temp-covid-growth",
    "sha": "2481beea34113c03c697e9dd8e02e6e2dd803446"
  },
  "prior": 0.35,
  "registered_at": "2026-11-12T22:49:54-0500",
  "registrant": "Claude (split-literature study, graph discovery)",
  "registration": {
    "alternatives": [
      "Behaviour: warm anomalies move people outdoors and cold ones indoors; this is part of the weather effect and is not separated from virus survival.",
      "Variant waves: growth is dominated by variant arrivals, which are unrelated to two-week weather anomalies and add noise but not bias.",
      "Reporting: hospitalisation definitions and testing changed over time (cases after 2022 are undercounted); year fixed effects and the use of growth rates absorb level changes only.",
      "Holidays: Thanksgiving to New Year gatherings fall in a season with specific anomalies in some years; harmonics remove the average seasonal pattern only."
    ],
    "graph_evidence": [
      {
        "concepts": [
          "temperature",
          "COVID-19"
        ],
        "direction": "split",
        "n_findings": 43,
        "n_studies": 29,
        "pubmed_ids": [
          32298883,
          32492607,
          32837322,
          32948946,
          33420694,
          33727663,
          33786420,
          32389157,
          32837623,
          33307083,
          34637125,
          36107871,
          34957431,
          37500034,
          35996423,
          34189495,
          34966171,
          35257688,
          34375443
        ],
        "relationship_id": "4FaXcuNJRt|DZBybMvxDI",
        "sign_split": {
          "minus": 17,
          "plus": 26,
          "sig_minus": 14,
          "sig_plus": 16
        },
        "summary": "The positive findings are mostly unadjusted Spearman or Pearson correlations between daily temperature and daily or cumulative case counts in one place over the first weeks or months of 2020, when cases and spring temperatures rose together (Jakarta, Oslo, India, Russia, Jordan, Korea, Pakistan). The negative findings mostly model transmission (reproduction number, growth, lagged incidence) with distributed lags, case-crossover designs or controls for government response (Wuhan and Hubei cities, four European countries, 153 countries above the turning point, US temperate regions, Mexico City, a meta-analysis). Four positive rows are not positive in substance: PMID 34189495 is '1 C decrease' and case fatality, PMID 34966171 is the relative risk at the 1st temperature percentile (cold raises transmissibility), PMID 35257688 says in its abstract that higher temperature goes with lower reproduction although the row is +0.875, and PMID 34375443 is body temperature (fever thresholds) grounded to 'temperature'. So the moderator is design and outcome: raw count correlations in a growth phase give either sign; models of transmission give negative."
      }
    ],
    "id": "split-temp-covid-growth",
    "kind": "confirmation",
    "literature": [],
    "plan": "Outcome: COVID-19 Daily Counts of Cases, Hospitalizations, and Deaths (rc75-m7u3), columns date_of_interest, case_count, hospitalized_count, all rows (2020-02-29 to 2025-10-13), through concord/socrata.py. Weeks are Monday to Sunday; H_t is the weekly sum of hospitalized_count; g_t = log(H_t / H_{t-1}); weeks from 2020-03-16 with H_t and H_{t-1} at least 20. Exposure: Central Park daily mean temperature (TMAX+TMIN)/2 minus its day-of-year normal (two-harmonic fit to 2010-2025), averaged over days 7 to 20 before the Thursday of week t. Model: OLS of g_t on the anomaly, g_{t-1}, year fixed effects and two annual harmonic pairs; Newey-West standard errors with 4 lags. Sensitivity, reported but not deciding: anomaly window days 0 to 13 and days 14 to 27; cases instead of admissions; dropping 2020; dropping the Omicron weeks (2021-12-13 to 2022-02-13). Secondary: for each calendar quarter from 2020 Q2 to 2025 Q3, Spearman correlation of daily mean temperature with daily case_count. Disclosure of what the project had seen before this registration. Graph side: all findings of both relationships and their PubMed abstracts were read (data/derived/graph/splits/findings.json, abstracts.json). City side: Central Park daily temperature (NOAA GHCN-Daily) has been used with shootings, heat-illness ED visits and EMS dispatches of type SHOT, STAB, DRUG, DRUGFC, EDP, EDPC, EDPM. For EMS dispatches (76xm-jjuj) the project has seen the count of every initial_call_type for calendar 2023 (ASTHMB 7,130; ASTHFC 2,813) and the date range of the file (2005-01-01 to 2026-09-30). No daily series of asthma dispatches has been queried by any agent (data/cache/socrata holds four 76xm-jjuj queries, none with an asthma code). For the COVID-19 daily counts (rc75-m7u3) I queried the column list and the first and last date (2020-02-29, 2025-10-13; 2,054 rows) and nothing else; no other agent has queried it.",
    "prediction": {
      "direction": "-",
      "magnitude": "Change in the weekly log growth rate of hospital admissions per 1 C of temperature anomaly averaged over days 7 to 20 before the week's midpoint: between -0.030 and -0.003. Raw quarterly Spearman correlations of daily mean temperature with daily confirmed cases: at least 25% of quarters significantly positive and at least 25% significantly negative (p < 0.05, ignoring autocorrelation, as the source studies did).",
      "moderator_prediction": "Secondary, scored separately and not deciding the verdict: among calendar quarters from 2020 Q2 to 2025 Q3, at least a quarter have a significantly positive raw correlation and at least a quarter a significantly negative one.",
      "refute_if": "The growth-rate coefficient is positive and its 95% interval excludes 0.",
      "support_if": "The growth-rate coefficient is negative and its 95% interval excludes 0."
    },
    "prior": 0.35,
    "registered_at": "2026-11-12T22:49:54-0500",
    "registrant": "Claude (split-literature study, graph discovery)",
    "sha256": "3ff45bbe83195bc2c6a1fc71ca14c5abc4b7edc1b47df41313f4370635d5e824",
    "statement": "In New York City, 2020 to 2025, weeks that follow warmer-than-normal weather for the time of year have slower growth of COVID-19 hospital admissions; and the raw correlation between daily temperature and daily case counts within single quarters, the design behind most positive findings in the graph, takes both signs depending on where the quarter falls in an epidemic wave."
  },
  "registration_commit": {
    "commit": "5443c2a34279deca1e9ec1a91e5c4ecb291a7832",
    "committed_at": "2026-11-12T22:49:54-05:00"
  },
  "registration_file_sha256": "e53eb02b752d3858d51aa82e9eed4239033e685d5a4a7301c4f09b7b0a594cfe",
  "registration_path": "claims/registered/split-temp-covid-growth.json",
  "registration_sha256": "3ff45bbe83195bc2c6a1fc71ca14c5abc4b7edc1b47df41313f4370635d5e824",
  "registration_text": "{\n \"id\": \"split-temp-covid-growth\",\n \"kind\": \"confirmation\",\n \"statement\": \"In New York City, 2020 to 2025, weeks that follow warmer-than-normal weather for the time of year have slower growth of COVID-19 hospital admissions; and the raw correlation between daily temperature and daily case counts within single quarters, the design behind most positive findings in the graph, takes both signs depending on where the quarter falls in an epidemic wave.\",\n \"graph_evidence\": [\n  {\n   \"relationship_id\": \"4FaXcuNJRt|DZBybMvxDI\",\n   \"concepts\": [\n    \"temperature\",\n    \"COVID-19\"\n   ],\n   \"n_findings\": 43,\n   \"n_studies\": 29,\n   \"sign_split\": {\n    \"plus\": 26,\n    \"minus\": 17,\n    \"sig_plus\": 16,\n    \"sig_minus\": 14\n   },\n   \"direction\": \"split\",\n   \"summary\": \"The positive findings are mostly unadjusted Spearman or Pearson correlations between daily temperature and daily or cumulative case counts in one place over the first weeks or months of 2020, when cases and spring temperatures rose together (Jakarta, Oslo, India, Russia, Jordan, Korea, Pakistan). The negative findings mostly model transmission (reproduction number, growth, lagged incidence) with distributed lags, case-crossover designs or controls for government response (Wuhan and Hubei cities, four European countries, 153 countries above the turning point, US temperate regions, Mexico City, a meta-analysis). Four positive rows are not positive in substance: PMID 34189495 is '1 C decrease' and case fatality, PMID 34966171 is the relative risk at the 1st temperature percentile (cold raises transmissibility), PMID 35257688 says in its abstract that higher temperature goes with lower reproduction although the row is +0.875, and PMID 34375443 is body temperature (fever thresholds) grounded to 'temperature'. So the moderator is design and outcome: raw count correlations in a growth phase give either sign; models of transmission give negative.\",\n   \"pubmed_ids\": [\n    32298883,\n    32492607,\n    32837322,\n    32948946,\n    33420694,\n    33727663,\n    33786420,\n    32389157,\n    32837623,\n    33307083,\n    34637125,\n    36107871,\n    34957431,\n    37500034,\n    35996423,\n    34189495,\n    34966171,\n    35257688,\n    34375443\n   ]\n  }\n ],\n \"literature\": [],\n \"prediction\": {\n  \"direction\": \"-\",\n  \"magnitude\": \"Change in the weekly log growth rate of hospital admissions per 1 C of temperature anomaly averaged over days 7 to 20 before the week's midpoint: between -0.030 and -0.003. Raw quarterly Spearman correlations of daily mean temperature with daily confirmed cases: at least 25% of quarters significantly positive and at least 25% significantly negative (p < 0.05, ignoring autocorrelation, as the source studies did).\",\n  \"support_if\": \"The growth-rate coefficient is negative and its 95% interval excludes 0.\",\n  \"refute_if\": \"The growth-rate coefficient is positive and its 95% interval excludes 0.\",\n  \"moderator_prediction\": \"Secondary, scored separately and not deciding the verdict: among calendar quarters from 2020 Q2 to 2025 Q3, at least a quarter have a significantly positive raw correlation and at least a quarter a significantly negative one.\"\n },\n \"plan\": \"Outcome: COVID-19 Daily Counts of Cases, Hospitalizations, and Deaths (rc75-m7u3), columns date_of_interest, case_count, hospitalized_count, all rows (2020-02-29 to 2025-10-13), through concord/socrata.py. Weeks are Monday to Sunday; H_t is the weekly sum of hospitalized_count; g_t = log(H_t / H_{t-1}); weeks from 2020-03-16 with H_t and H_{t-1} at least 20. Exposure: Central Park daily mean temperature (TMAX+TMIN)/2 minus its day-of-year normal (two-harmonic fit to 2010-2025), averaged over days 7 to 20 before the Thursday of week t. Model: OLS of g_t on the anomaly, g_{t-1}, year fixed effects and two annual harmonic pairs; Newey-West standard errors with 4 lags. Sensitivity, reported but not deciding: anomaly window days 0 to 13 and days 14 to 27; cases instead of admissions; dropping 2020; dropping the Omicron weeks (2021-12-13 to 2022-02-13). Secondary: for each calendar quarter from 2020 Q2 to 2025 Q3, Spearman correlation of daily mean temperature with daily case_count. Disclosure of what the project had seen before this registration. Graph side: all findings of both relationships and their PubMed abstracts were read (data/derived/graph/splits/findings.json, abstracts.json). City side: Central Park daily temperature (NOAA GHCN-Daily) has been used with shootings, heat-illness ED visits and EMS dispatches of type SHOT, STAB, DRUG, DRUGFC, EDP, EDPC, EDPM. For EMS dispatches (76xm-jjuj) the project has seen the count of every initial_call_type for calendar 2023 (ASTHMB 7,130; ASTHFC 2,813) and the date range of the file (2005-01-01 to 2026-09-30). No daily series of asthma dispatches has been queried by any agent (data/cache/socrata holds four 76xm-jjuj queries, none with an asthma code). For the COVID-19 daily counts (rc75-m7u3) I queried the column list and the first and last date (2020-02-29, 2025-10-13; 2,054 rows) and nothing else; no other agent has queried it.\",\n \"alternatives\": [\n  \"Behaviour: warm anomalies move people outdoors and cold ones indoors; this is part of the weather effect and is not separated from virus survival.\",\n  \"Variant waves: growth is dominated by variant arrivals, which are unrelated to two-week weather anomalies and add noise but not bias.\",\n  \"Reporting: hospitalisation definitions and testing changed over time (cases after 2022 are undercounted); year fixed effects and the use of growth rates absorb level changes only.\",\n  \"Holidays: Thanksgiving to New Year gatherings fall in a season with specific anomalies in some years; harmonics remove the average seasonal pattern only.\"\n ],\n \"prior\": 0.35,\n \"registrant\": \"Claude (split-literature study, graph discovery)\",\n \"registered_at\": \"2026-11-12T22:49:54-0500\",\n \"sha256\": \"3ff45bbe83195bc2c6a1fc71ca14c5abc4b7edc1b47df41313f4370635d5e824\"\n}",
  "statement": "In New York City, 2020 to 2025, weeks that follow warmer-than-normal weather for the time of year have slower growth of COVID-19 hospital admissions; and the raw correlation between daily temperature and daily case counts within single quarters, the design behind most positive findings in the graph, takes both signs depending on where the quarter falls in an epidemic wave."
}

History in this collection

  1. addedv1.0.0Nov 14, 2026this version