{
  "id": 170986,
  "title": "Distribution of last 3 FVC Readings",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/170986",
  "author_name": "",
  "post_date": "2020-07-29T21:39:48.968005400Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1968272%2F21348bf7e8cab20af76fb073b8be1956%2FFVC.png?generation=1596058630010298&amp;alt=media\" alt=\"\"></p>\n\n<p>Had a look at the distribution of the last 3 FVC readings vs. the rest and there's definitely a distribution shift (as we might expect as patients generally worsen rather than improve in the dataset). Though perhaps not as large as I might have expected.</p>\n\n<p>So this suggests we might be able to bias our solutions towards the 'last 3' distribution. </p>",
  "messages": [
    {
      "id": "951090",
      "postDate": "07/29/2020 21:39:48",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1968272%2F21348bf7e8cab20af76fb073b8be1956%2FFVC.png?generation=1596058630010298&amp;alt=media\" alt=\"\"></p>\n\n<p>Had a look at the distribution of the last 3 FVC readings vs. the rest and there's definitely a distribution shift (as we might expect as patients generally worsen rather than improve in the dataset). Though perhaps not as large as I might have expected.</p>\n\n<p>So this suggests we might be able to bias our solutions towards the 'last 3' distribution. </p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1968272%2F21348bf7e8cab20af76fb073b8be1956%2FFVC.png?generation=1596058630010298&amp;alt=media)\n\nHad a look at the distribution of the last 3 FVC readings vs. the rest and there's definitely a distribution shift (as we might expect as patients generally worsen rather than improve in the dataset). Though perhaps not as large as I might have expected.\n\nSo this suggests we might be able to bias our solutions towards the 'last 3' distribution.",
      "votes": null
    },
    {
      "id": "951109",
      "postDate": "07/29/2020 22:10:39",
      "content": "<p>Yeah, the way the scoring works in the competition (only taking the last 3 measurements into account) has some interesting consequenses when estimating the confidences. (I have a half-finished notebook about a related topic I should try and finish some day =)</p>",
      "rawMarkdown": "Yeah, the way the scoring works in the competition (only taking the last 3 measurements into account) has some interesting consequenses when estimating the confidences. (I have a half-finished notebook about a related topic I should try and finish some day =)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 951109,
      "author_name": "christoffer",
      "author_url": "",
      "post_date": "07/29/2020 22:10:39",
      "content": "<p>Yeah, the way the scoring works in the competition (only taking the last 3 measurements into account) has some interesting consequenses when estimating the confidences. (I have a half-finished notebook about a related topic I should try and finish some day =)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "951090": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1968272%2F21348bf7e8cab20af76fb073b8be1956%2FFVC.png?generation=1596058630010298&amp;alt=media)\n\nHad a look at the distribution of the last 3 FVC readings vs. the rest and there's definitely a distribution shift (as we might expect as patients generally worsen rather than improve in the dataset). Though perhaps not as large as I might have expected.\n\nSo this suggests we might be able to bias our solutions towards the 'last 3' distribution.",
    "951109": "Yeah, the way the scoring works in the competition (only taking the last 3 measurements into account) has some interesting consequenses when estimating the confidences. (I have a half-finished notebook about a related topic I should try and finish some day =)"
  },
  "source": "meta"
}