{
  "id": 2386,
  "title": "Small bug correction to evaluation metric",
  "url": "/competitions/MerckActivity/discussion/2386",
  "author_name": "",
  "post_date": "2012-08-18T22:59:09.570Z",
  "votes": null,
  "comment_count": 4,
  "views": 2930,
  "content": "<p>Hi Everyone,</p>\r\n<p>We discovered a small bug in the evaluation metric as implemented for the website. We've corrected it and rescored the leaderboard. The sample script for R-squared is not affected by this change.</p>\r\n<p>Cheers,</p>\r\n<p>Joyce</p>",
  "messages": [
    {
      "id": "13202",
      "postDate": "08/18/2012 22:59:09",
      "content": "<p>Hi Everyone,</p>\r\n<p>We discovered a small bug in the evaluation metric as implemented for the website. We've corrected it and rescored the leaderboard. The sample script for R-squared is not affected by this change.</p>\r\n<p>Cheers,</p>\r\n<p>Joyce</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "13356",
      "postDate": "08/23/2012 13:11:37",
      "content": "<p>Re: Sample R-Squared Script</p>\r\n<p>Your sample script is a bit complicated. A simpler solution is:</p>\r\n<p>RSquared &lt;- cor(x,y)^2</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "14064",
      "postDate": "09/07/2012 16:25:22",
      "content": "<p>The leaderboard says &quot;Data Set Weighted Correlation Coefficient.&quot; Are the weights the same for every dataset (e.g. the leaderboard score is the average R2 for each dataset)?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "14067",
      "postDate": "09/07/2012 16:31:17",
      "content": "<p>I think it uses a single R2 calculated from your submission, doesn't it?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "14069",
      "postDate": "09/07/2012 16:46:20",
      "content": "<p>[quote=Zach;14064]</p>\r\n<p>The leaderboard says &quot;Data Set Weighted Correlation Coefficient.&quot; Are the weights the same for every dataset (e.g. the leaderboard score is the average R2 for each dataset)?</p>\r\n<p>[/quote]</p>\r\n<p>It's an average.</p>\r\n<p>In the evaluation page:</p>\r\n<p><a href=\"https://www.kaggle.com/c/MerckActivity/details/evaluation\">https://www.kaggle.com/c/MerckActivity/details/evaluation</a></p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 13356,
      "author_name": "biasvariance",
      "author_url": "",
      "post_date": "08/23/2012 13:11:37",
      "content": "<p>Re: Sample R-Squared Script</p>\r\n<p>Your sample script is a bit complicated. A simpler solution is:</p>\r\n<p>RSquared &lt;- cor(x,y)^2</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 14064,
      "author_name": "zachmayer",
      "author_url": "",
      "post_date": "09/07/2012 16:25:22",
      "content": "<p>The leaderboard says &quot;Data Set Weighted Correlation Coefficient.&quot; Are the weights the same for every dataset (e.g. the leaderboard score is the average R2 for each dataset)?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 14067,
      "author_name": "benoitplante",
      "author_url": "",
      "post_date": "09/07/2012 16:31:17",
      "content": "<p>I think it uses a single R2 calculated from your submission, doesn't it?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 14069,
      "author_name": "blindape",
      "author_url": "",
      "post_date": "09/07/2012 16:46:20",
      "content": "<p>[quote=Zach;14064]</p>\r\n<p>The leaderboard says &quot;Data Set Weighted Correlation Coefficient.&quot; Are the weights the same for every dataset (e.g. the leaderboard score is the average R2 for each dataset)?</p>\r\n<p>[/quote]</p>\r\n<p>It's an average.</p>\r\n<p>In the evaluation page:</p>\r\n<p><a href=\"https://www.kaggle.com/c/MerckActivity/details/evaluation\">https://www.kaggle.com/c/MerckActivity/details/evaluation</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "13202": "",
    "13356": "",
    "14064": "",
    "14067": "",
    "14069": ""
  },
  "source": "meta"
}