Pardon if this is a stupid question. In the test dataset we got, there is only one FVC reading and the CT scan to make future predictions. Is this a feature for all further data coming along? Or in the validation and future test set are we going to have multiple FVC readings or is it definitely just 1 FVC reading?
", "messages": [ { "id": 932137, "postDate": "2020-07-16T18:24:25.783Z", "content": "Pardon if this is a stupid question. In the test dataset we got, there is only one FVC reading and the CT scan to make future predictions. Is this a feature for all further data coming along? Or in the validation and future test set are we going to have multiple FVC readings or is it definitely just 1 FVC reading?
", "rawMarkdown": "Pardon if this is a stupid question. In the test dataset we got, there is only one FVC reading and the CT scan to make future predictions. Is this a feature for all further data coming along? Or in the validation and future test set are we going to have multiple FVC readings or is it definitely just 1 FVC reading?", "votes": 3 }, { "id": 932459, "postDate": "2020-07-17T04:43:11.143Z", "content": "Thank you so much! Also, do you think its a good idea to make groups of two in the training, considering we only have only 1 FVC and 1 CT scan for testing. As apposed to making it time series problem, would it work if we instead make each training example 1 CT and 1 FVC. This way we get many more training examples and also simulate the actual test set.(Our target variables would be every FVC reading after the one taken for training,also very much like the test set).
", "rawMarkdown": "Thank you so much! Also, do you think its a good idea to make groups of two in the training, considering we only have only 1 FVC and 1 CT scan for testing. As apposed to making it time series problem, would it work if we instead make each training example 1 CT and 1 FVC. This way we get many more training examples and also simulate the actual test set.(Our target variables would be every FVC reading after the one taken for training,also very much like the test set).", "votes": 1, "replies": [ { "id": 932766, "postDate": "2020-07-17T09:06:48.657Z", "content": "Worth a try!
\nMost important is to first get some baseline up for running for FVC and/or CT. Then you can experiment with this approach and others, whether it's valid or not
Not a stupid question at all.
\nDefinitely one FVC measurement. And one CT scan. CT scan taken at Weeks=0. FVC could be any other week, but only the first FVC is given (doesn't have to correspond to Weeks=0 like the CT scan)
Will do! thanks!
", "rawMarkdown": "Will do! thanks!" } ], "comments": [ { "id": 932459, "author_name": "mseb", "author_url": "", "post_date": "2020-07-17T04:43:11.143000", "content": "Thank you so much! Also, do you think its a good idea to make groups of two in the training, considering we only have only 1 FVC and 1 CT scan for testing. As apposed to making it time series problem, would it work if we instead make each training example 1 CT and 1 FVC. This way we get many more training examples and also simulate the actual test set.(Our target variables would be every FVC reading after the one taken for training,also very much like the test set).
", "votes": 1, "replies": [ { "id": 932766, "author_name": "Rafi Hai", "author_url": "", "post_date": "2020-07-17T09:06:48.657000", "content": "Worth a try!
\nMost important is to first get some baseline up for running for FVC and/or CT. Then you can experiment with this approach and others, whether it's valid or not
Not a stupid question at all.
\nDefinitely one FVC measurement. And one CT scan. CT scan taken at Weeks=0. FVC could be any other week, but only the first FVC is given (doesn't have to correspond to Weeks=0 like the CT scan)
Will do! thanks!
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "932137": "Pardon if this is a stupid question. In the test dataset we got, there is only one FVC reading and the CT scan to make future predictions. Is this a feature for all further data coming along? Or in the validation and future test set are we going to have multiple FVC readings or is it definitely just 1 FVC reading?", "932459": "Thank you so much! Also, do you think its a good idea to make groups of two in the training, considering we only have only 1 FVC and 1 CT scan for testing. As apposed to making it time series problem, would it work if we instead make each training example 1 CT and 1 FVC. This way we get many more training examples and also simulate the actual test set.(Our target variables would be every FVC reading after the one taken for training,also very much like the test set).", "932272": "Not a stupid question at all. \nDefinitely one FVC measurement. And one CT scan. CT scan taken at Weeks=0. FVC could be any other week, but only the first FVC is given (doesn't have to correspond to Weeks=0 like the CT scan)", "932457": "", "932456": "", "932871": "Will do! thanks!" } }