{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreigklusvctacubqvfu2y36475zyxui2zvhamnq5l5e5gaauwkllqei",
    "uri": "at://did:plc:wwyqal4cnqhuwyacdj7rqq3n/app.bsky.feed.post/3mljr233uazx2"
  },
  "path": "/t/dichotomization/26337?page=4#post_79",
  "publishedAt": "2026-05-10T13:47:02.000Z",
  "site": "https://discourse.datamethods.org",
  "tags": [
    "https://www.sciencedirect.com/science/article/pii/S106345842501283X",
    "[16]"
  ],
  "textContent": "Maybe some progress (?)\n\nTurkiewicz A et al. Why Most Responder Analyses are Misleading. Osteoarthritis and Cartilage; 34(5), May 2026: 716-720.\n\nhttps://www.sciencedirect.com/science/article/pii/S106345842501283X\n\n> _“Importantly, it is not the existence of pre-specified thresholds for clinical relevance, such as minimally important difference, that is the problem, even though categorising continuous scales is almost always a bad idea[16]. The problem arises when this threshold is applied to individual observed change from baseline, because observed change from baseline is a suboptimal representation of a person’s response to treatment when dealing with outcomes that naturally fluctuate…”_",
  "title": "Dichotomization"
}