{
  "id": 2218,
  "title": "how is the evaluation score aggregated across groups?",
  "url": "/competitions/Raising-Money-to-Fund-an-Organizational-Mission/discussion/2218",
  "author_name": "",
  "post_date": "2012-07-20T22:39:39.187Z",
  "votes": null,
  "comment_count": 1,
  "views": 1342,
  "content": "<p>I think I understand how the &quot;Average of Top 75%&quot; score works ... you drop my lowest scored 25% in each group, then average donation^1.14 across the highest scored 75% in each group.&nbsp; But there are 134 distinct groups in the test set, with group sizes ranging\r\n from 280,480 rows all the way down to just 158 rows.&nbsp; How is the Average of Top 75% for each group aggregated across the 134 groups to calculate the evaluation score?</p>",
  "messages": [
    {
      "id": "12516",
      "postDate": "07/20/2012 22:39:39",
      "content": "<p>I think I understand how the &quot;Average of Top 75%&quot; score works ... you drop my lowest scored 25% in each group, then average donation^1.14 across the highest scored 75% in each group.&nbsp; But there are 134 distinct groups in the test set, with group sizes ranging\r\n from 280,480 rows all the way down to just 158 rows.&nbsp; How is the Average of Top 75% for each group aggregated across the 134 groups to calculate the evaluation score?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "12517",
      "postDate": "07/20/2012 22:50:21",
      "content": "<p>It's just taking those top 75% in each group, throwing them together, and doing a straight average. R code is here, if that helps: https://www.kaggle.com/wiki/AverageAmongTopP.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 12517,
      "author_name": "dchudz",
      "author_url": "",
      "post_date": "07/20/2012 22:50:21",
      "content": "<p>It's just taking those top 75% in each group, throwing them together, and doing a straight average. R code is here, if that helps: https://www.kaggle.com/wiki/AverageAmongTopP.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "12516": "",
    "12517": ""
  },
  "source": "meta"
}