{
  "id": 19012,
  "title": "Median vs. Mean",
  "url": "/competitions/second-annual-data-science-bowl/discussion/19012",
  "author_name": "",
  "post_date": "2016-02-16T22:30:19.580Z",
  "votes": null,
  "comment_count": 1,
  "views": 482,
  "content": "<p>Most scripts take the average for different sax/stack predictions for the same patient. I was originally thinking that a median function would be more robust against outliers. However, at least for me the performance for a median function was significantly worse. Does anyone have an intuition why this might be the case? </p>",
  "messages": [
    {
      "id": "108370",
      "postDate": "02/16/2016 22:30:19",
      "content": "<p>Most scripts take the average for different sax/stack predictions for the same patient. I was originally thinking that a median function would be more robust against outliers. However, at least for me the performance for a median function was significantly worse. Does anyone have an intuition why this might be the case? </p>",
      "rawMarkdown": "Most scripts take the average for different sax/stack predictions for the same patient. I was originally thinking that a median function would be more robust against outliers. However, at least for me the performance for a median function was significantly worse. Does anyone have an intuition why this might be the case?",
      "votes": null
    },
    {
      "id": "108480",
      "postDate": "02/17/2016 17:51:08",
      "content": "<p>For some pixels the heart is present less than half of the sax/stacks. </p>",
      "rawMarkdown": "For some pixels the heart is present less than half of the sax/stacks.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 108480,
      "author_name": "woolsey",
      "author_url": "",
      "post_date": "02/17/2016 17:51:08",
      "content": "<p>For some pixels the heart is present less than half of the sax/stacks. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "108370": "Most scripts take the average for different sax/stack predictions for the same patient. I was originally thinking that a median function would be more robust against outliers. However, at least for me the performance for a median function was significantly worse. Does anyone have an intuition why this might be the case?",
    "108480": "For some pixels the heart is present less than half of the sax/stacks."
  },
  "source": "meta"
}