{
  "id": 192332,
  "title": "multiple answers with same timestamp",
  "url": "/competitions/riiid-test-answer-prediction/discussion/192332",
  "author_name": "",
  "post_date": "2020-10-21T02:27:01.189695300Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>This yields three records for the same user/task &amp; bundle.</p>\n<p><code>train_df[train_df.row_id.isin([5533153,5533154,5533155])]</code></p>\n<p>Question:</p>\n<ul>\n<li>Does the user answer three questions simultaneously?</li>\n<li>Is there another attribute I can use to sort this correctly (by time/order of presentation/answering?)</li>\n</ul>\n<p>I think it was already established that task ID is not ordered. Maybe row_id can be used?</p>",
  "messages": [
    {
      "id": "1055685",
      "postDate": "10/21/2020 02:27:01",
      "content": "<p>This yields three records for the same user/task &amp; bundle.</p>\n<p><code>train_df[train_df.row_id.isin([5533153,5533154,5533155])]</code></p>\n<p>Question:</p>\n<ul>\n<li>Does the user answer three questions simultaneously?</li>\n<li>Is there another attribute I can use to sort this correctly (by time/order of presentation/answering?)</li>\n</ul>\n<p>I think it was already established that task ID is not ordered. Maybe row_id can be used?</p>",
      "rawMarkdown": "This yields three records for the same user/task & bundle.\n\n```train_df[train_df.row_id.isin([5533153,5533154,5533155])]```\n\nQuestion:\n\n- Does the user answer three questions simultaneously?\n- Is there another attribute I can use to sort this correctly (by time/order of presentation/answering?)\n\nI think it was already established that task ID is not ordered. Maybe row_id can be used?",
      "votes": null
    },
    {
      "id": "1056529",
      "postDate": "10/21/2020 19:41:21",
      "content": "<p>The repeated timestamps have the same <code>task_container_id</code> but different <code>user_answers</code> which seems to suggest they are are two point/multi select questions. Their different <code>content_id</code>s also seem to support this. Now there is also <code>task_container_id</code>, which seems to be monotonically increasing, except it isn't. As you can read in this <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465\" target=\"_blank\">discussion post</a>, <code>task_container_id</code> captures when a user first sees a task, whereas <code>timestamp</code> captures the order in which a user actually does the task. So it would seem that users are free to start a task, go to another, and finish the second before the first.</p>",
      "rawMarkdown": "The repeated timestamps have the same `task_container_id` but different `user_answers` which seems to suggest they are are two point/multi select questions. Their different `content_id`s also seem to support this. Now there is also `task_container_id`, which seems to be monotonically increasing, except it isn't. As you can read in this [discussion post](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465), `task_container_id` captures when a user first sees a task, whereas `timestamp` captures the order in which a user actually does the task. So it would seem that users are free to start a task, go to another, and finish the second before the first.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1056529,
      "author_name": "tuckerarrants",
      "author_url": "",
      "post_date": "10/21/2020 19:41:21",
      "content": "<p>The repeated timestamps have the same <code>task_container_id</code> but different <code>user_answers</code> which seems to suggest they are are two point/multi select questions. Their different <code>content_id</code>s also seem to support this. Now there is also <code>task_container_id</code>, which seems to be monotonically increasing, except it isn't. As you can read in this <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465\" target=\"_blank\">discussion post</a>, <code>task_container_id</code> captures when a user first sees a task, whereas <code>timestamp</code> captures the order in which a user actually does the task. So it would seem that users are free to start a task, go to another, and finish the second before the first.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1055685": "This yields three records for the same user/task & bundle.\n\n```train_df[train_df.row_id.isin([5533153,5533154,5533155])]```\n\nQuestion:\n\n- Does the user answer three questions simultaneously?\n- Is there another attribute I can use to sort this correctly (by time/order of presentation/answering?)\n\nI think it was already established that task ID is not ordered. Maybe row_id can be used?",
    "1056529": "The repeated timestamps have the same `task_container_id` but different `user_answers` which seems to suggest they are are two point/multi select questions. Their different `content_id`s also seem to support this. Now there is also `task_container_id`, which seems to be monotonically increasing, except it isn't. As you can read in this [discussion post](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465), `task_container_id` captures when a user first sees a task, whereas `timestamp` captures the order in which a user actually does the task. So it would seem that users are free to start a task, go to another, and finish the second before the first."
  },
  "source": "meta"
}