I came across 289 sessions in which one of the 22 levels is not present.
\nThe levels not present are 7, 15, 20 and 21.
\nIs this an error in the data?
I came across 289 sessions in which one of the 22 levels is not present.
\nThe levels not present are 7, 15, 20 and 21.
\nIs this an error in the data?
Hi @shashwatraman,
\nIn the beginning, I assume the student uses save code to skip some specific levels. However, it's weird that the elapsed_time difference between the last event in level l and the first event in level l+1 is short. For example, session 20100018421509572 skips level 15 and the elapsed_time difference between the last event in level 14 (i.e., \"observation_click\") and the first one in level 16 (i.e., \"navigate_click\") is only 1581ms.
Hence, I think it could result from some unexpected data record missing issues. Also, the data errors in columns index and elapsed_time might be relevant, which is discussed here.
Hope this helps, thanks.
", "rawMarkdown": "Hi @shashwatraman, \n\nIn the beginning, I assume the student uses **save code** to skip some specific levels. However, it's weird that the `elapsed_time` difference between the last event in level *l* and the first event in level *l+1* is **short**. For example, session `20100018421509572` skips level 15 and the `elapsed_time` difference between the last event in level 14 (*i.e.,* \"observation_click\") and the first one in level 16 (*i.e.,* \"navigate_click\") is only 1581ms.\n\nHence, I think it could result from some unexpected data record missing issues. Also, the data errors in columns `index` and `elapsed_time` might be relevant, which is discussed [here](https://www.kaggle.com/competitions/predict-student-performance-from-game-play/discussion/384342#2134312).\n\nHope this helps, thanks.", "votes": null }, { "id": "2159355", "postDate": "02/25/2023 17:44:34", "content": "Hi @abaojiang,
\nYes, I also observed the short difference. But in some sessions the difference is big too. Nevertheless, are you dropping these sessions, or just ignoring them?
Thanks for replying.
", "rawMarkdown": "Hi @abaojiang,\nYes, I also observed the short difference. But in some sessions the difference is big too. Nevertheless, are you dropping these sessions, or just ignoring them?\n\nThanks for replying.", "votes": null }, { "id": "2159361", "postDate": "02/25/2023 17:48:52", "content": "Hi @shashwatraman,
\nSo far, I just keep all the sessions. For those with missing levels, I pad special tokens to align with normal sessions!
", "rawMarkdown": "Hi @shashwatraman,\n\nSo far, I just keep all the sessions. For those with missing levels, I pad **special tokens** to align with normal sessions!", "votes": null }, { "id": "2159376", "postDate": "02/25/2023 18:03:20", "content": "Thanks for the information.
", "rawMarkdown": "Thanks for the information.", "votes": null }, { "id": "2174184", "postDate": "03/08/2023 23:11:13", "content": "Thanks for raising this. We can confirm that the gameplay sessions in the training data do not use save codes. This is likely an error in the way that the game records the data for these levels, as you noted.
", "rawMarkdown": "Thanks for raising this. We can confirm that the gameplay sessions in the training data do not use save codes. This is likely an error in the way that the game records the data for these levels, as you noted.", "votes": null }, { "id": "2174524", "postDate": "03/09/2023 07:36:47", "content": "Thanks for the clarification @alexmlfranklin . Can you disclose if also in the test set all the gameplay sessions do not use save codes ?
", "rawMarkdown": "Thanks for the clarification @alexmlfranklin . Can you disclose if also in the test set all the gameplay sessions do not use save codes ?", "votes": null }, { "id": "2175290", "postDate": "03/09/2023 18:53:29", "content": "Yes, this is also the case for the test set.
", "rawMarkdown": "Yes, this is also the case for the test set.", "votes": null } ], "comments": [ { "id": 2159238, "author_name": "abaojiang", "author_url": "", "post_date": "02/25/2023 15:31:10", "content": "Hi @shashwatraman,
\nIn the beginning, I assume the student uses save code to skip some specific levels. However, it's weird that the elapsed_time difference between the last event in level l and the first event in level l+1 is short. For example, session 20100018421509572 skips level 15 and the elapsed_time difference between the last event in level 14 (i.e., \"observation_click\") and the first one in level 16 (i.e., \"navigate_click\") is only 1581ms.
Hence, I think it could result from some unexpected data record missing issues. Also, the data errors in columns index and elapsed_time might be relevant, which is discussed here.
Hope this helps, thanks.
", "votes": null, "replies": [ { "id": 2159355, "author_name": "shashwatraman", "author_url": "", "post_date": "02/25/2023 17:44:34", "content": "Hi @abaojiang,
\nYes, I also observed the short difference. But in some sessions the difference is big too. Nevertheless, are you dropping these sessions, or just ignoring them?
Thanks for replying.
", "votes": null, "replies": [ { "id": 2159361, "author_name": "abaojiang", "author_url": "", "post_date": "02/25/2023 17:48:52", "content": "Hi @shashwatraman,
\nSo far, I just keep all the sessions. For those with missing levels, I pad special tokens to align with normal sessions!
", "votes": null, "replies": [ { "id": 2159376, "author_name": "shashwatraman", "author_url": "", "post_date": "02/25/2023 18:03:20", "content": "Thanks for the information.
", "votes": null, "replies": [] } ] } ] } ] }, { "id": 2174184, "author_name": "alexmlfranklin", "author_url": "", "post_date": "03/08/2023 23:11:13", "content": "Thanks for raising this. We can confirm that the gameplay sessions in the training data do not use save codes. This is likely an error in the way that the game records the data for these levels, as you noted.
", "votes": null, "replies": [ { "id": 2174524, "author_name": "steubk", "author_url": "", "post_date": "03/09/2023 07:36:47", "content": "Thanks for the clarification @alexmlfranklin . Can you disclose if also in the test set all the gameplay sessions do not use save codes ?
", "votes": null, "replies": [ { "id": 2175290, "author_name": "alexmlfranklin", "author_url": "", "post_date": "03/09/2023 18:53:29", "content": "Yes, this is also the case for the test set.
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2158801": "I came across **289 sessions** in which one of the 22 levels is not present. \nThe levels not present are 7, 15, 20 and 21.\nIs this an error in the data?", "2159238": "Hi @shashwatraman, \n\nIn the beginning, I assume the student uses **save code** to skip some specific levels. However, it's weird that the `elapsed_time` difference between the last event in level *l* and the first event in level *l+1* is **short**. For example, session `20100018421509572` skips level 15 and the `elapsed_time` difference between the last event in level 14 (*i.e.,* \"observation_click\") and the first one in level 16 (*i.e.,* \"navigate_click\") is only 1581ms.\n\nHence, I think it could result from some unexpected data record missing issues. Also, the data errors in columns `index` and `elapsed_time` might be relevant, which is discussed [here](https://www.kaggle.com/competitions/predict-student-performance-from-game-play/discussion/384342#2134312).\n\nHope this helps, thanks.", "2159355": "Hi @abaojiang,\nYes, I also observed the short difference. But in some sessions the difference is big too. Nevertheless, are you dropping these sessions, or just ignoring them?\n\nThanks for replying.", "2159361": "Hi @shashwatraman,\n\nSo far, I just keep all the sessions. For those with missing levels, I pad **special tokens** to align with normal sessions!", "2159376": "Thanks for the information.", "2174184": "Thanks for raising this. We can confirm that the gameplay sessions in the training data do not use save codes. This is likely an error in the way that the game records the data for these levels, as you noted.", "2174524": "Thanks for the clarification @alexmlfranklin . Can you disclose if also in the test set all the gameplay sessions do not use save codes ?", "2175290": "Yes, this is also the case for the test set." }, "source": "meta" }