{
  "id": 5138,
  "title": "train4 target answer is incorrect?",
  "url": "/competitions/cause-effect-pairs/discussion/5138",
  "author_name": "",
  "post_date": "2013-07-18T12:26:05.660Z",
  "votes": null,
  "comment_count": 4,
  "views": 1214,
  "content": "<p>The&nbsp;target table answer to the train4 dataset is:&nbsp; A caused B.</p>\r\n<p>The correct answer is that, from the data, we cannot tell that A caused B or B caused A.</p>\r\n<p>Therefore the target table target answer should be corrected to 0.</p>\r\n<p>Am I correct?</p>\r\n<p>Note:&nbsp; I view the 6 to 20 outliers as outliers, especially since they are a small proportion of the total (&lt;= 20/802 pairs).</p>\r\n<p>Even _with_ these outliers, there would either be:</p>\r\n<p>{a relation between A and B without causality, or</p>\r\n<p>A and B are consequences of a common cause, or</p>\r\n<p>A and B are independent}.</p>\r\n<p>Comments please!</p>\r\n<p></p>\r\n<p></p>",
  "messages": [
    {
      "id": "27405",
      "postDate": "07/18/2013 12:26:05",
      "content": "<p>The&nbsp;target table answer to the train4 dataset is:&nbsp; A caused B.</p>\r\n<p>The correct answer is that, from the data, we cannot tell that A caused B or B caused A.</p>\r\n<p>Therefore the target table target answer should be corrected to 0.</p>\r\n<p>Am I correct?</p>\r\n<p>Note:&nbsp; I view the 6 to 20 outliers as outliers, especially since they are a small proportion of the total (&lt;= 20/802 pairs).</p>\r\n<p>Even _with_ these outliers, there would either be:</p>\r\n<p>{a relation between A and B without causality, or</p>\r\n<p>A and B are consequences of a common cause, or</p>\r\n<p>A and B are independent}.</p>\r\n<p>Comments please!</p>\r\n<p></p>\r\n<p></p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "27407",
      "postDate": "07/18/2013 12:35:45",
      "content": "<p>Hi, We're not going to know if it's incorrect or not. Same with your other post. If you think they're incorrect so be it, but it makes no difference to us as competitors. There's some level of noise in the data so expect some labels to be incorrect</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "27411",
      "postDate": "07/18/2013 14:34:20",
      "content": "<p>[quote=Domcastro;27407]</p>\r\n<p>Hi, We're not going to know if it's incorrect or not. Same with your other post. If you think they're incorrect so be it, but it makes no difference to us as competitors. There's some level of noise in the data so expect some labels to be incorrect</p>\r\n<p>[/quote]</p>\r\n<p>I am well aware of the noise in the data;&nbsp; That is not a full discussion or explanation of why the &quot;labels&quot; are incorrect.</p>\r\n<p>Let's hold the job of Data Scientist with high regard.&nbsp; These are not just &quot;labels&quot;.&nbsp; These target answers represent &quot;Truth&quot; in this competition.&nbsp; The customer will often be interested in the true nature of their data, and not be in it for just sport.&nbsp; Let's\r\n be true to a mission statement of Kaggle:</p>\r\n<p>&quot;Kaggle provides cutting-edge data science results to companies of all sizes. We have a proven track-record of solving real-world problems across a diverse array of industries including life sciences, financial services, energy, information technology, and\r\n retail&quot;.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "27417",
      "postDate": "07/18/2013 15:23:40",
      "content": "<p>The labels are nothing to do with Kaggle - they are provded by the &quot;customer&quot;, ie the experts in the areas.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "27559",
      "postDate": "07/23/2013 17:14:23",
      "content": "<p>Both train2 and train4 are artificial pairs so there is no doubt about the true causal direction. However, this does not mean that it is possible to infer it from data (because the noise, the finite sample and the particular functional relationship). This is part of the problem. I will not answer more questions about pairs of variables.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 27407,
      "author_name": "domcastro",
      "author_url": "",
      "post_date": "07/18/2013 12:35:45",
      "content": "<p>Hi, We're not going to know if it's incorrect or not. Same with your other post. If you think they're incorrect so be it, but it makes no difference to us as competitors. There's some level of noise in the data so expect some labels to be incorrect</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 27411,
      "author_name": "kevinh7777",
      "author_url": "",
      "post_date": "07/18/2013 14:34:20",
      "content": "<p>[quote=Domcastro;27407]</p>\r\n<p>Hi, We're not going to know if it's incorrect or not. Same with your other post. If you think they're incorrect so be it, but it makes no difference to us as competitors. There's some level of noise in the data so expect some labels to be incorrect</p>\r\n<p>[/quote]</p>\r\n<p>I am well aware of the noise in the data;&nbsp; That is not a full discussion or explanation of why the &quot;labels&quot; are incorrect.</p>\r\n<p>Let's hold the job of Data Scientist with high regard.&nbsp; These are not just &quot;labels&quot;.&nbsp; These target answers represent &quot;Truth&quot; in this competition.&nbsp; The customer will often be interested in the true nature of their data, and not be in it for just sport.&nbsp; Let's\r\n be true to a mission statement of Kaggle:</p>\r\n<p>&quot;Kaggle provides cutting-edge data science results to companies of all sizes. We have a proven track-record of solving real-world problems across a diverse array of industries including life sciences, financial services, energy, information technology, and\r\n retail&quot;.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 27417,
      "author_name": "domcastro",
      "author_url": "",
      "post_date": "07/18/2013 15:23:40",
      "content": "<p>The labels are nothing to do with Kaggle - they are provded by the &quot;customer&quot;, ie the experts in the areas.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 27559,
      "author_name": "iguyon",
      "author_url": "",
      "post_date": "07/23/2013 17:14:23",
      "content": "<p>Both train2 and train4 are artificial pairs so there is no doubt about the true causal direction. However, this does not mean that it is possible to infer it from data (because the noise, the finite sample and the particular functional relationship). This is part of the problem. I will not answer more questions about pairs of variables.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "27405": "",
    "27407": "",
    "27411": "",
    "27417": "",
    "27559": ""
  },
  "source": "meta"
}