{
  "id": 182971,
  "title": "Is there any correlation between the \"Percent\" in training data and \"Confidence\" in submission data?",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/182971",
  "author_name": "",
  "post_date": "2020-09-15T04:14:34.521616Z",
  "votes": 3,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Is there any correlation between the \"Percent\" in training data and \"Confidence\" in submission data?<br>\nCould we use \"Percent\" as the label of \"Confidence\" ?</p>\n<p>Many thanks!</p>",
  "messages": [
    {
      "id": "1010788",
      "postDate": "09/15/2020 04:14:34",
      "content": "<p>Is there any correlation between the \"Percent\" in training data and \"Confidence\" in submission data?<br>\nCould we use \"Percent\" as the label of \"Confidence\" ?</p>\n<p>Many thanks!</p>",
      "rawMarkdown": "Is there any correlation between the \"Percent\" in training data and \"Confidence\" in submission data?\nCould we use \"Percent\" as the label of \"Confidence\" ?\n\nMany thanks!",
      "votes": null
    },
    {
      "id": "1011031",
      "postDate": "09/15/2020 07:57:11",
      "content": "<p>'Percent' means the percentage of the predicted FVC that a patient has. So, e.g. if a patient should have (= what a standard formula predicts for a healthy patient with same sex, height, race and age) a FVC of 3000 and the patient's measured FVC is 2700, then Percent = 90. So, from that perspective, it is not directly telling you something about the confidence (e.g. it is not some kind of label giving some kind of information about confidence in the value).</p>\n<p>Could it indirectly contain useful information about variation/how easy it is to predict (e.g. at lower levels of Percent our confidence might also be lower - or higher - , because somehow results tend to vary more at some level of Percent), yes, of course, that might be the case.</p>",
      "rawMarkdown": "'Percent' means the percentage of the predicted FVC that a patient has. So, e.g. if a patient should have (= what a standard formula predicts for a healthy patient with same sex, height, race and age) a FVC of 3000 and the patient's measured FVC is 2700, then Percent = 90. So, from that perspective, it is not directly telling you something about the confidence (e.g. it is not some kind of label giving some kind of information about confidence in the value).\n\nCould it indirectly contain useful information about variation/how easy it is to predict (e.g. at lower levels of Percent our confidence might also be lower - or higher - , because somehow results tend to vary more at some level of Percent), yes, of course, that might be the case.",
      "votes": null
    },
    {
      "id": "1011484",
      "postDate": "09/15/2020 14:06:08",
      "content": "<p>Thanks a lot! It solved my problem!</p>",
      "rawMarkdown": "Thanks a lot! It solved my problem!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1011031,
      "author_name": "bjoernholzhauer",
      "author_url": "",
      "post_date": "09/15/2020 07:57:11",
      "content": "<p>'Percent' means the percentage of the predicted FVC that a patient has. So, e.g. if a patient should have (= what a standard formula predicts for a healthy patient with same sex, height, race and age) a FVC of 3000 and the patient's measured FVC is 2700, then Percent = 90. So, from that perspective, it is not directly telling you something about the confidence (e.g. it is not some kind of label giving some kind of information about confidence in the value).</p>\n<p>Could it indirectly contain useful information about variation/how easy it is to predict (e.g. at lower levels of Percent our confidence might also be lower - or higher - , because somehow results tend to vary more at some level of Percent), yes, of course, that might be the case.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1011484,
      "author_name": "crissallan",
      "author_url": "",
      "post_date": "09/15/2020 14:06:08",
      "content": "<p>Thanks a lot! It solved my problem!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1010788": "Is there any correlation between the \"Percent\" in training data and \"Confidence\" in submission data?\nCould we use \"Percent\" as the label of \"Confidence\" ?\n\nMany thanks!",
    "1011031": "'Percent' means the percentage of the predicted FVC that a patient has. So, e.g. if a patient should have (= what a standard formula predicts for a healthy patient with same sex, height, race and age) a FVC of 3000 and the patient's measured FVC is 2700, then Percent = 90. So, from that perspective, it is not directly telling you something about the confidence (e.g. it is not some kind of label giving some kind of information about confidence in the value).\n\nCould it indirectly contain useful information about variation/how easy it is to predict (e.g. at lower levels of Percent our confidence might also be lower - or higher - , because somehow results tend to vary more at some level of Percent), yes, of course, that might be the case.",
    "1011484": "Thanks a lot! It solved my problem!"
  },
  "source": "meta"
}