Like the example here: https://www.kaggle.com/code/metric/mean-weighted-columnwise-aucroc. I am having difficulty calculating the metrics. The results I obtain are quite different from those on the public leaderboard.
", "messages": [ { "id": 3309760, "postDate": "2025-11-01T10:28:59.230Z", "content": "We use the wrapper for scikit-learn’s r2_score. You can find the metric script here
", "rawMarkdown": "We use the wrapper for scikit-learn’s [r2_score](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.r2_score.html). You can find the metric script [here](https://www.kaggle.com/code/metric/r2-score?scriptVersionId=151169785)", "votes": 3, "replies": [ { "id": 3311326, "postDate": "2025-11-04T18:23:14.100Z", "content": "So, a log transfomation is not included in the R-squared calculations?\nI ask because the \"Estimating Pasture…\" paper's section \"4.5 Evaluation Protocol\" says that \"a log-stabilizing transformation is applied to all target variables before computing R-squared values: y_trans = log(1+y)\". Was it decided not to do this after all?
", "rawMarkdown": "So, a log transfomation is not included in the R-squared calculations?\nI ask because the [\"Estimating Pasture...\"](https://arxiv.org/abs/2510.22916) paper's section \"4.5 Evaluation Protocol\" says that \"a log-stabilizing transformation is applied to all target variables before computing R-squared values: y_trans = log(1+y)\". Was it decided not to do this after all?", "replies": [ { "id": 3311332, "postDate": "2025-11-04T18:36:31.170Z", "rawMarkdown": "", "isDeleted": true }, { "id": 3311346, "postDate": "2025-11-04T19:18:32.593Z", "content": "I understand that the Kaggle pages are the official source ✅ 🙂 . . . \nfyi, the paper is listed in the Citation on the Data page.\nI just wanted to check, in case it was an oversight and the competition host, @qiyuliaocsiro , really wanted to use the log(1+y) transformation. Thanks.
", "rawMarkdown": "I understand that the Kaggle pages are the official source ✅ 🙂 . . . \nfyi, the paper is listed in the Citation on the Data page.\nI just wanted to check, in case it was an oversight and the competition host, @qiyuliaocsiro , really wanted to use the log(1+y) transformation. Thanks." } ] }, { "id": 3311355, "postDate": "2025-11-04T19:52:33.440Z", "content": "@martynaplomecka , Sorry for my chatter here: I was reading the discussion posts in Recently Posted order so I had not yet gotten to the other metric discussion posts and the resolution of the scoring metric, clearly given on the Overview page. Thank you.
", "rawMarkdown": "@martynaplomecka , Sorry for my chatter here: I was reading the discussion posts in Recently Posted order so I had not yet gotten to the other metric discussion posts and the resolution of the **scoring metric, clearly given on the Overview page**. Thank you.", "votes": 1 } ] }, { "id": 3309658, "postDate": "2025-11-01T06:00:23.353Z", "content": "Like the example here: https://www.kaggle.com/code/metric/mean-weighted-columnwise-aucroc. I am having difficulty calculating the metrics. The results I obtain are quite different from those on the public leaderboard.
", "rawMarkdown": "Like the example here: https://www.kaggle.com/code/metric/mean-weighted-columnwise-aucroc. I am having difficulty calculating the metrics. The results I obtain are quite different from those on the public leaderboard.", "votes": 2 } ], "comments": [ { "id": 3309760, "author_name": "MartynaPlomecka", "author_url": "", "post_date": "2025-11-01T10:28:59.230000", "content": "We use the wrapper for scikit-learn’s r2_score. You can find the metric script here
", "votes": 3, "replies": [ { "id": 3311326, "author_name": "Daniel Dewey", "author_url": "", "post_date": "2025-11-04T18:23:14.100000", "content": "So, a log transfomation is not included in the R-squared calculations?\nI ask because the \"Estimating Pasture…\" paper's section \"4.5 Evaluation Protocol\" says that \"a log-stabilizing transformation is applied to all target variables before computing R-squared values: y_trans = log(1+y)\". Was it decided not to do this after all?
", "votes": 0, "replies": [ { "id": 3311332, "author_name": "", "author_url": "", "post_date": "2025-11-04T18:36:31.170000", "content": "", "votes": 0, "replies": [] }, { "id": 3311346, "author_name": "Daniel Dewey", "author_url": "", "post_date": "2025-11-04T19:18:32.593000", "content": "I understand that the Kaggle pages are the official source ✅ 🙂 . . . \nfyi, the paper is listed in the Citation on the Data page.\nI just wanted to check, in case it was an oversight and the competition host, @qiyuliaocsiro , really wanted to use the log(1+y) transformation. Thanks.
", "votes": 0, "replies": [] } ] }, { "id": 3311355, "author_name": "Daniel Dewey", "author_url": "", "post_date": "2025-11-04T19:52:33.440000", "content": "@martynaplomecka , Sorry for my chatter here: I was reading the discussion posts in Recently Posted order so I had not yet gotten to the other metric discussion posts and the resolution of the scoring metric, clearly given on the Overview page. Thank you.
", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "3309760": "We use the wrapper for scikit-learn’s [r2_score](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.r2_score.html). You can find the metric script [here](https://www.kaggle.com/code/metric/r2-score?scriptVersionId=151169785)", "3309658": "Like the example here: https://www.kaggle.com/code/metric/mean-weighted-columnwise-aucroc. I am having difficulty calculating the metrics. The results I obtain are quite different from those on the public leaderboard." } }