{
  "alt": "Table with one row for each cutoff from {fact:cut_lo} to {fact:cut_hi} counties, giving the number of states and counties, the number of states with an interval below zero and above zero, and four p-values for equal slopes: points weighted by births, points with each county counting once, proportion weighted by births and proportion with each county counting once. The p-values for proportion are large in every row.",
  "caption": "Read across a row for one cutoff. The count of states with an interval below zero barely moves, no interval lies above zero at any cutoff, and the p-values for proportional slopes stay above {fact:p_prop_min} in every row. In points the p-value depends on the weights: it is small for every cutoff above {fact:cut_lo} counties when each county counts once and larger when counties are weighted by births.",
  "data": [
    {
      "above": 0,
      "below": 11,
      "counties": 518,
      "cutoff": 3,
      "p_points_unweighted": "0.264",
      "p_points_weighted": "0.109",
      "p_prop_unweighted": "0.784",
      "p_prop_weighted": "0.788",
      "states": 41
    },
    {
      "above": 0,
      "below": 11,
      "counties": 506,
      "cutoff": 4,
      "p_points_unweighted": "0.006",
      "p_points_weighted": "0.121",
      "p_prop_unweighted": "0.710",
      "p_prop_weighted": "0.790",
      "states": 37
    },
    {
      "above": 0,
      "below": 11,
      "counties": 498,
      "cutoff": 5,
      "p_points_unweighted": "0.006",
      "p_points_weighted": "0.118",
      "p_prop_unweighted": "0.648",
      "p_prop_weighted": "0.743",
      "states": 35
    },
    {
      "above": 0,
      "below": 10,
      "counties": 473,
      "cutoff": 6,
      "p_points_unweighted": "0.006",
      "p_points_weighted": "0.117",
      "p_prop_unweighted": "0.344",
      "p_prop_weighted": "0.639",
      "states": 30
    },
    {
      "above": 0,
      "below": 10,
      "counties": 449,
      "cutoff": 8,
      "p_points_unweighted": "0.009",
      "p_points_weighted": "0.108",
      "p_prop_unweighted": "0.343",
      "p_prop_weighted": "0.664",
      "states": 26
    },
    {
      "above": 0,
      "below": 9,
      "counties": 423,
      "cutoff": 10,
      "p_points_unweighted": "0.0005",
      "p_points_weighted": "0.165",
      "p_prop_unweighted": "0.319",
      "p_prop_weighted": "0.601",
      "states": 23
    }
  ],
  "id": "cutoff_sensitivity",
  "notes": [
    "A p-value tests that every state in the row has the same slope. It is the share of wild-bootstrap draws (replicates and seed in the methods note) in which Cochran's Q is at least as large as observed, so the smallest value it can take is one over the number of draws plus one.",
    "Large counties are the counties that CDC WONDER names individually: those with {fact:wonder_threshold} residents or more. WONDER pools the births of every smaller county in a state into one row, so smaller counties are not in these data. Of the {fact:n_named} large counties named in every year from 2018 to 2024, California ({fact:n_ca}) and {fact:n_dropped} counties with a withheld smoking count are left out, which leaves {fact:n_complete}.",
    "County median household income is the County Health Rankings 2024 release, a Census Small Area Income and Poverty Estimates value for 2022. Its year differs from the births, which are from 2018 to 2024."
  ],
  "source": "CDC WONDER natality, tobacco use during pregnancy, births in 2018-2024; full counts of births; counties ranked on County Health Rankings 2024 median household income (Census SAIPE 2022); state slopes weighted by births or unweighted as labeled; p-values from a wild bootstrap",
  "spec": {
    "columns": [
      {
        "key": "cutoff",
        "label": "Fewest complete counties"
      },
      {
        "key": "states",
        "label": "States"
      },
      {
        "key": "counties",
        "label": "Counties"
      },
      {
        "key": "below",
        "label": "States with interval below zero"
      },
      {
        "key": "above",
        "label": "States with interval above zero"
      },
      {
        "key": "p_points_weighted",
        "label": "p, points, weighted"
      },
      {
        "key": "p_points_unweighted",
        "label": "p, points, each county once"
      },
      {
        "key": "p_prop_weighted",
        "label": "p, proportion, weighted"
      },
      {
        "key": "p_prop_unweighted",
        "label": "p, proportion, each county once"
      }
    ]
  },
  "subtitle": "States with a slope of their own, states with an interval below or above zero, and wild-bootstrap p-values for equal slopes in percentage points and in proportion, by the fewest complete counties a state needs, 2018 to 2024",
  "title": "Raising the minimum number of counties from {fact:cut_lo} to {fact:cut_hi} leaves {fact:below_min} to {fact:below_max} states with an interval below zero and none above it, and the test of equal proportional slopes never rejects",
  "type": "table",
  "width": "wide"
}