For each patient, we are given his info and for each week he goes for test the CT scan (Dicom images ) are given and for training label we are given the FCV and percent (confidence)
\nIn the test data, we are given the patients info and Dicom images and we are asked to predict the FCV and percent(confidence level)
\nIs my understanding correct?
Another doubt: How is confidence level calculated?
\nfrom my understanding, the confidence level (say 95%)tells that if I take the N samples from the population and build a 95% confidence interval on each sample then 95% percent of the samples will have the population mean between the confidence interval.
\n Is this definition anyhow related to the calculation of confidence here?
For each patient, we are given his info and for each week he goes for test the CT scan (Dicom images ) are given and for training label we are given the FCV and percent (confidence)
\nIn the test data, we are given the patients info and Dicom images and we are asked to predict the FCV and percent(confidence level)
\nIs my understanding correct?
Another doubt: How is confidence level calculated?
\nfrom my understanding, the confidence level (say 95%)tells that if I take the N samples from the population and build a 95% confidence interval on each sample then 95% percent of the samples will have the population mean between the confidence interval.
\n Is this definition anyhow related to the calculation of confidence here?
Hi. We are given baseline information, a single baseline CT (week 0), and subsequent FVC measurements are taken (other weeks). Although asked to predict FVCs for all weeks (-12 to 133) in reality only \"the last 3 visits\" (which are unknown to us) are going to be evaluated.
\nI think the term confidence is unfortunate and is an overloading of the term. It's confusing because a lower \"confidence\" value actually means that the model is \"more confident\" about its prediction.
\nGiven the metric, other users have shown the relationship between \"confidence\" and predicted FVC that will give the optimal score.
\nUsers in their notebooks, particularly those using Quantile Regression, often calculate the \"confidence\" as the difference between the upper value and the lower value.
", "rawMarkdown": "Hi. We are given baseline information, a single baseline CT (week 0), and subsequent FVC measurements are taken (other weeks). Although asked to predict FVCs for all weeks (-12 to 133) in reality only \"the last 3 visits\" (which are unknown to us) are going to be evaluated.\n\nI think the term confidence is unfortunate and is an overloading of the term. It's confusing because a lower \"confidence\" value actually means that the model is \"more confident\" about its prediction.\n\nGiven the metric, other users have [shown the relationship between \"confidence\" and predicted FVC](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/173636) that will give the optimal score. \n\nUsers in their notebooks, particularly those using Quantile Regression, often calculate the \"confidence\" as the difference between the upper value and the lower value.", "votes": null }, { "id": "981962", "postDate": "08/22/2020 22:18:46", "content": "No, there is only one CT scan per patient. The CT scan is considered the week 0, weeks before are negatives. But the patients keep revising the doctor and doing FVC tests, which shows how many air their lungs are capable to blow (the disease keeps damaging the lungs). In the test, we need to use both image and the FVC tests to predict the FVC and confidence interval of our predictions.
", "rawMarkdown": "No, there is only one CT scan per patient. The CT scan is considered the week 0, weeks before are negatives. But the patients keep revising the doctor and doing FVC tests, which shows how many air their lungs are capable to blow (the disease keeps damaging the lungs). In the test, we need to use both image and the FVC tests to predict the FVC and confidence interval of our predictions.", "votes": null } ], "comments": [ { "id": 981925, "author_name": "jjinho", "author_url": "", "post_date": "08/22/2020 21:10:10", "content": "Hi. We are given baseline information, a single baseline CT (week 0), and subsequent FVC measurements are taken (other weeks). Although asked to predict FVCs for all weeks (-12 to 133) in reality only \"the last 3 visits\" (which are unknown to us) are going to be evaluated.
\nI think the term confidence is unfortunate and is an overloading of the term. It's confusing because a lower \"confidence\" value actually means that the model is \"more confident\" about its prediction.
\nGiven the metric, other users have shown the relationship between \"confidence\" and predicted FVC that will give the optimal score.
\nUsers in their notebooks, particularly those using Quantile Regression, often calculate the \"confidence\" as the difference between the upper value and the lower value.
", "votes": null, "replies": [] }, { "id": 981962, "author_name": "cafalchio", "author_url": "", "post_date": "08/22/2020 22:18:46", "content": "No, there is only one CT scan per patient. The CT scan is considered the week 0, weeks before are negatives. But the patients keep revising the doctor and doing FVC tests, which shows how many air their lungs are capable to blow (the disease keeps damaging the lungs). In the test, we need to use both image and the FVC tests to predict the FVC and confidence interval of our predictions.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "981895": "For each patient, we are given his info and for each week he goes for test the CT scan (Dicom images ) are given and for training label we are given the FCV and percent (confidence)\nIn the test data, we are given the patients info and Dicom images and we are asked to predict the FCV and percent(confidence level)\nIs my understanding correct?\n\n**Another doubt: How is confidence level calculated?**\nfrom my understanding, the confidence level (say 95%)tells that if I take the N samples from the population and build a 95% confidence interval on each sample then 95% percent of the samples will have the population mean between the confidence interval. \n Is this definition anyhow related to the calculation of confidence here?", "981925": "Hi. We are given baseline information, a single baseline CT (week 0), and subsequent FVC measurements are taken (other weeks). Although asked to predict FVCs for all weeks (-12 to 133) in reality only \"the last 3 visits\" (which are unknown to us) are going to be evaluated.\n\nI think the term confidence is unfortunate and is an overloading of the term. It's confusing because a lower \"confidence\" value actually means that the model is \"more confident\" about its prediction.\n\nGiven the metric, other users have [shown the relationship between \"confidence\" and predicted FVC](https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/173636) that will give the optimal score. \n\nUsers in their notebooks, particularly those using Quantile Regression, often calculate the \"confidence\" as the difference between the upper value and the lower value.", "981962": "No, there is only one CT scan per patient. The CT scan is considered the week 0, weeks before are negatives. But the patients keep revising the doctor and doing FVC tests, which shows how many air their lungs are capable to blow (the disease keeps damaging the lungs). In the test, we need to use both image and the FVC tests to predict the FVC and confidence interval of our predictions." }, "source": "meta" }