{
  "id": 171353,
  "title": "Doubts about \"Percent\" column ",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/171353",
  "author_name": "",
  "post_date": "2020-07-31T13:49:39.618582900Z",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hi everyone,\nI have seen many notebook that use percent as a training feature (not baseline percent), so I'd like to know if this column is available also for test data or only baseline percent is available</p>",
  "messages": [
    {
      "id": "953064",
      "postDate": "07/31/2020 13:49:39",
      "content": "<p>Hi everyone,\nI have seen many notebook that use percent as a training feature (not baseline percent), so I'd like to know if this column is available also for test data or only baseline percent is available</p>",
      "rawMarkdown": "Hi everyone,\nI have seen many notebook that use percent as a training feature (not baseline percent), so I'd like to know if this column is available also for test data or only baseline percent is available",
      "votes": null
    },
    {
      "id": "953194",
      "postDate": "07/31/2020 15:40:14",
      "content": "<p>You have the baseline <code>FVC</code> and <code>Percent</code> of the patients in test set.</p>",
      "rawMarkdown": "You have the baseline `FVC` and `Percent` of the patients in test set.",
      "votes": null
    },
    {
      "id": "953210",
      "postDate": "07/31/2020 15:49:05",
      "content": "<p>thanks for reply :). So we have to predict FVC of other weeks from baseline Percent and baseline FVC only (excluding sex, age ..)? Therefore using the week percent () in training is not representative of test set situation, right?</p>",
      "rawMarkdown": "thanks for reply :). So we have to predict FVC of other weeks from baseline Percent and baseline FVC only (excluding sex, age ..)? Therefore using the week percent () in training is not representative of test set situation, right?",
      "votes": null
    },
    {
      "id": "953221",
      "postDate": "07/31/2020 16:00:02",
      "content": "<p>You're welcome. You don't have to exclude any feature especially <code>Age</code>, <code>Sex</code> and <code>SmokingStatus</code>, but I wouldn't use <code>Percent</code> in the models because test set only have the baseline value.</p>",
      "rawMarkdown": "You're welcome. You don't have to exclude any feature especially `Age`, `Sex` and `SmokingStatus`, but I wouldn't use `Percent` in the models because test set only have the baseline value.",
      "votes": null
    },
    {
      "id": "953231",
      "postDate": "07/31/2020 16:06:22",
      "content": "<p>Thanks again, I also think like you.\nThe doubt arose because I have seen many notebooks that use \"percent\". So I thought about asking for clarification on his presence or not in test set.\nUltimately, I think like you, it's probably better not to use percent different from baseline :)</p>",
      "rawMarkdown": "Thanks again, I also think like you.\nThe doubt arose because I have seen many notebooks that use \"percent\". So I thought about asking for clarification on his presence or not in test set.\nUltimately, I think like you, it's probably better not to use percent different from baseline :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 953194,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "07/31/2020 15:40:14",
      "content": "<p>You have the baseline <code>FVC</code> and <code>Percent</code> of the patients in test set.</p>",
      "votes": null,
      "replies": [
        {
          "id": 953210,
          "author_name": "agostinodorano",
          "author_url": "",
          "post_date": "07/31/2020 15:49:05",
          "content": "<p>thanks for reply :). So we have to predict FVC of other weeks from baseline Percent and baseline FVC only (excluding sex, age ..)? Therefore using the week percent () in training is not representative of test set situation, right?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 953221,
          "author_name": "gunesevitan",
          "author_url": "",
          "post_date": "07/31/2020 16:00:02",
          "content": "<p>You're welcome. You don't have to exclude any feature especially <code>Age</code>, <code>Sex</code> and <code>SmokingStatus</code>, but I wouldn't use <code>Percent</code> in the models because test set only have the baseline value.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 953231,
          "author_name": "agostinodorano",
          "author_url": "",
          "post_date": "07/31/2020 16:06:22",
          "content": "<p>Thanks again, I also think like you.\nThe doubt arose because I have seen many notebooks that use \"percent\". So I thought about asking for clarification on his presence or not in test set.\nUltimately, I think like you, it's probably better not to use percent different from baseline :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "953064": "Hi everyone,\nI have seen many notebook that use percent as a training feature (not baseline percent), so I'd like to know if this column is available also for test data or only baseline percent is available",
    "953194": "You have the baseline `FVC` and `Percent` of the patients in test set.",
    "953210": "thanks for reply :). So we have to predict FVC of other weeks from baseline Percent and baseline FVC only (excluding sex, age ..)? Therefore using the week percent () in training is not representative of test set situation, right?",
    "953221": "You're welcome. You don't have to exclude any feature especially `Age`, `Sex` and `SmokingStatus`, but I wouldn't use `Percent` in the models because test set only have the baseline value.",
    "953231": "Thanks again, I also think like you.\nThe doubt arose because I have seen many notebooks that use \"percent\". So I thought about asking for clarification on his presence or not in test set.\nUltimately, I think like you, it's probably better not to use percent different from baseline :)"
  },
  "source": "meta"
}