{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibwzrerhbbf6lblcpw7zahht2q7sigpg7a5pmiyhaa4le7cjsgfw4",
    "uri": "at://did:plc:wwyqal4cnqhuwyacdj7rqq3n/app.bsky.feed.post/3mfcu42wwxc22"
  },
  "path": "/t/good-reading-on-when-doubly-robust-is-worth-the-effort/17033#post_9",
  "publishedAt": "2026-02-18T12:32:47.000Z",
  "site": "https://discourse.datamethods.org",
  "textContent": "It does to me at least.\n\nI would like to always see a comparison with an optimal method that doesn’t use weights in any way. I find that weights are always ad hoc and there’s often a full modeling approach that relies on the log likelihood.",
  "title": "Good reading on when \"doubly robust\" is worth the effort?"
}