{
  "id": 12558,
  "title": "Quadratic weighted kappa",
  "url": "/competitions/diabetic-retinopathy-detection/discussion/12558",
  "author_name": "",
  "post_date": "2015-02-18T16:36:04.140Z",
  "votes": 13,
  "comment_count": 6,
  "views": 6579,
  "content": "<p>For those who are&nbsp;interested to know more about kappa...see links below</p>\n\n<p>http://en.wikipedia.org/wiki/Cohen's_kappa</p>\n\n<p>http://www.medcalc.org/manual/kappa.php</p>\n\n<p>http://vassarstats.net/kappa.html</p>",
  "messages": [
    {
      "id": "64512",
      "postDate": "02/18/2015 16:36:04",
      "content": "<p>For those who are&nbsp;interested to know more about kappa...see links below</p>\n\n<p>http://en.wikipedia.org/wiki/Cohen's_kappa</p>\n\n<p>http://www.medcalc.org/manual/kappa.php</p>\n\n<p>http://vassarstats.net/kappa.html</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64527",
      "postDate": "02/18/2015 20:25:56",
      "content": "<p>It's not immediately clear to me why all zeros produces a LB score of 0.00 (as well as a submission of all 1's, etc.)</p>\n<p>Going to have to work through the math, I guess.</p>\n<p>Perhaps the metric doesn't go below zero? (Even though it says it does on the evaluation page.)</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64529",
      "postDate": "02/18/2015 21:08:41",
      "content": "<p>I have seen the metric go below zero...used random guesses between 0 to 4 and submitted once. &nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64642",
      "postDate": "02/20/2015 13:01:10",
      "content": "<p>@9811107 me too</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64651",
      "postDate": "02/20/2015 16:57:52",
      "content": "<p>This might be helpful</p>\n\n<p>https://www.kaggle.com/c/asap-aes/forums/t/1289/scoring-metric-verification</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "470852",
      "postDate": "02/13/2019 17:21:49",
      "content": "<p>That's very strange for me. I would have expected a negative Kappa for all 0 and all 1 and a positive one for 2. I'm really interested of your results if you do the math.</p>",
      "rawMarkdown": "That's very strange for me. I would have expected a negative Kappa for all 0 and all 1 and a positive one for 2. I'm really interested of your results if you do the math.",
      "votes": null
    },
    {
      "id": "470859",
      "postDate": "02/13/2019 17:31:28",
      "content": "<p>Actually I think I have the answer. If all the adoption categories have the same number of observations in the test file, you should indeed have 0.</p>",
      "rawMarkdown": "Actually I think I have the answer. If all the adoption categories have the same number of observations in the test file, you should indeed have 0.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 64527,
      "author_name": "inversion",
      "author_url": "",
      "post_date": "02/18/2015 20:25:56",
      "content": "<p>It's not immediately clear to me why all zeros produces a LB score of 0.00 (as well as a submission of all 1's, etc.)</p>\n<p>Going to have to work through the math, I guess.</p>\n<p>Perhaps the metric doesn't go below zero? (Even though it says it does on the evaluation page.)</p>",
      "votes": null,
      "replies": [
        {
          "id": 470852,
          "author_name": "jpmary",
          "author_url": "",
          "post_date": "02/13/2019 17:21:49",
          "content": "<p>That's very strange for me. I would have expected a negative Kappa for all 0 and all 1 and a positive one for 2. I'm really interested of your results if you do the math.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 470859,
          "author_name": "jpmary",
          "author_url": "",
          "post_date": "02/13/2019 17:31:28",
          "content": "<p>Actually I think I have the answer. If all the adoption categories have the same number of observations in the test file, you should indeed have 0.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 64529,
      "author_name": "",
      "author_url": "",
      "post_date": "02/18/2015 21:08:41",
      "content": "<p>I have seen the metric go below zero...used random guesses between 0 to 4 and submitted once. &nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64642,
      "author_name": "jaymoore",
      "author_url": "",
      "post_date": "02/20/2015 13:01:10",
      "content": "<p>@9811107 me too</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64651,
      "author_name": "numb3rs",
      "author_url": "",
      "post_date": "02/20/2015 16:57:52",
      "content": "<p>This might be helpful</p>\n\n<p>https://www.kaggle.com/c/asap-aes/forums/t/1289/scoring-metric-verification</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "64512": "",
    "64527": "",
    "64529": "",
    "64642": "",
    "64651": "",
    "470852": "That's very strange for me. I would have expected a negative Kappa for all 0 and all 1 and a positive one for 2. I'm really interested of your results if you do the math.",
    "470859": "Actually I think I have the answer. If all the adoption categories have the same number of observations in the test file, you should indeed have 0."
  },
  "source": "meta"
}