![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3617078%2Fb9f9423f5006fb5766ade4b8effcf7f5%2F2020-09-06%2012.16.01.png?generation=1599366034401839&alt=media\")
From the data description, we can know that \"Percent\" feature is \" a computed field which approximates the patient's FVC as a percent of the typical FVC for a person of similar \"
\nSo basically, the people with similar condition should have similar typical FVC, right?
\n
\nI try to calculate the typical FVC and try to find some useful features from it, then I noticed something weird.
\nThe way I calculated the typical FVC : typical FVC = FVC / (perecnt / 100 ).
\nAs you can see from picture below, the people with similar condition ( age : 58, gender : male, SmokingStatus : Ex-smoker) have a very huge different typical FVC between first and second red box.
So in my understanding, there are might be three possible reasons:
\nWhat do you think ? Or did I miss something?
\n
From the data description, we can know that \"Percent\" feature is \" a computed field which approximates the patient's FVC as a percent of the typical FVC for a person of similar \"
\nSo basically, the people with similar condition should have similar typical FVC, right?
\n
\nI try to calculate the typical FVC and try to find some useful features from it, then I noticed something weird.
\nThe way I calculated the typical FVC : typical FVC = FVC / (perecnt / 100 ).
\nAs you can see from picture below, the people with similar condition ( age : 58, gender : male, SmokingStatus : Ex-smoker) have a very huge different typical FVC between first and second red box.
So in my understanding, there are might be three possible reasons:
\nWhat do you think ? Or did I miss something?
\n
The percentage is a calculated value that is determined from the expected full volume. The full volume is a function of the age, sex, race (Caucasian, Hispanic, Black) and height. You can google and find some on-line calculators.
\nSo you will see differences because you don't know the race or height. If you include 'typical_FVC' in your model than your indirectly including race and height.
", "rawMarkdown": "The percentage is a calculated value that is determined from the expected full volume. The full volume is a function of the age, sex, race (Caucasian, Hispanic, Black) and height. You can google and find some on-line calculators.\n\nSo you will see differences because you don't know the race or height. If you include 'typical_FVC' in your model than your indirectly including race and height.", "votes": null }, { "id": "999958", "postDate": "09/06/2020 06:55:50", "content": "Thank you so much for the information!! No wonder there are lots of people trying to get the \"height\" information.
", "rawMarkdown": "Thank you so much for the information!! No wonder there are lots of people trying to get the \"height\" information.", "votes": null }, { "id": "999977", "postDate": "09/06/2020 07:14:38", "content": "https://www.cdc.gov/niosh/topics/spirometry/refcalculator.html
\nHere's link to calculator. Note that the calculated values are in liters and our data is in milliliters.
", "rawMarkdown": "https://www.cdc.gov/niosh/topics/spirometry/refcalculator.html\n\nHere's link to calculator. Note that the calculated values are in liters and our data is in milliliters.", "votes": null }, { "id": "999989", "postDate": "09/06/2020 07:28:42", "content": "Thanks a lot for the kind sharing and remind. Good luck in private score :)
", "rawMarkdown": "Thanks a lot for the kind sharing and remind. Good luck in private score :)", "votes": null }, { "id": "1000265", "postDate": "09/06/2020 12:22:14", "content": "super work
", "rawMarkdown": "super work", "votes": null }, { "id": "1000317", "postDate": "09/06/2020 13:11:34", "content": "Unfortunately, my model didn't improve even if adding predicted height.
\nOne of the reasons for this may be no information about patient's race.
Good findings, do you further try if it improves your CV? To me, the feature is not improving my CV.
", "rawMarkdown": "Good findings, do you further try if it improves your CV? To me, the feature is not improving my CV.", "votes": null }, { "id": "1012971", "postDate": "09/16/2020 12:34:06", "content": "In fact, no. Basically it has similar information as percent of min weeks.
", "rawMarkdown": "In fact, no. Basically it has similar information as percent of min weeks.", "votes": null } ], "comments": [ { "id": 999952, "author_name": "pcjimmmy", "author_url": "", "post_date": "09/06/2020 06:47:00", "content": "The percentage is a calculated value that is determined from the expected full volume. The full volume is a function of the age, sex, race (Caucasian, Hispanic, Black) and height. You can google and find some on-line calculators.
\nSo you will see differences because you don't know the race or height. If you include 'typical_FVC' in your model than your indirectly including race and height.
", "votes": null, "replies": [ { "id": 999958, "author_name": "xiejialun", "author_url": "", "post_date": "09/06/2020 06:55:50", "content": "Thank you so much for the information!! No wonder there are lots of people trying to get the \"height\" information.
", "votes": null, "replies": [] }, { "id": 999977, "author_name": "pcjimmmy", "author_url": "", "post_date": "09/06/2020 07:14:38", "content": "https://www.cdc.gov/niosh/topics/spirometry/refcalculator.html
\nHere's link to calculator. Note that the calculated values are in liters and our data is in milliliters.
", "votes": null, "replies": [] }, { "id": 999989, "author_name": "xiejialun", "author_url": "", "post_date": "09/06/2020 07:28:42", "content": "Thanks a lot for the kind sharing and remind. Good luck in private score :)
", "votes": null, "replies": [] }, { "id": 1000317, "author_name": "drtausamaru", "author_url": "", "post_date": "09/06/2020 13:11:34", "content": "Unfortunately, my model didn't improve even if adding predicted height.
\nOne of the reasons for this may be no information about patient's race.
Good findings, do you further try if it improves your CV? To me, the feature is not improving my CV.
", "votes": null, "replies": [ { "id": 1012971, "author_name": "xiejialun", "author_url": "", "post_date": "09/16/2020 12:34:06", "content": "In fact, no. Basically it has similar information as percent of min weeks.
", "votes": null, "replies": [] } ] }, { "id": 1000265, "author_name": "vijaysimhareddyp", "author_url": "", "post_date": "09/06/2020 12:22:14", "content": "super work
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "999835": "From the data description, we can know that \"Percent\" feature is \" a computed field which approximates the patient's FVC as a percent of the typical FVC for a person of similar \"\nSo basically, the people with similar condition should have similar typical FVC, right? \n