{ "id": 186585, "title": "Figuring out how to predict FVC and Sigma", "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/186585", "author_name": "", "post_date": "2020-09-25T04:29:52.543169500Z", "votes": 3, "comment_count": 2, "views": 0, "content": "

Hello,

\n

I've been really struggling to understand how we are supposed to predict sigma. I have a working neural network that predicts FVC just fine but I can't figure out how we predict sigma. I've read a couple posts on in the discussions and some notebooks but the implementations haven't been very clear to me.
\nThe best I could surmise was that we need to predict FVC in different quantiles subtract the first and third ones to get some kind of avg.

\n

A couple of the popular ones that people have been passing around on the forums are:
\nhttps://www.kaggle.com/ulrich07/osic-multiple-quantile-regression-starter/notebook
\nhttps://www.kaggle.com/yasufuminakama/osic-lgb-baseline

\n

It would be really helpful to have just a basic explanation of how we predict sigma.

\n

Thank you!

", "messages": [ { "id": "1026092", "postDate": "09/25/2020 04:29:52", "content": "

Hello,

\n

I've been really struggling to understand how we are supposed to predict sigma. I have a working neural network that predicts FVC just fine but I can't figure out how we predict sigma. I've read a couple posts on in the discussions and some notebooks but the implementations haven't been very clear to me.
\nThe best I could surmise was that we need to predict FVC in different quantiles subtract the first and third ones to get some kind of avg.

\n

A couple of the popular ones that people have been passing around on the forums are:
\nhttps://www.kaggle.com/ulrich07/osic-multiple-quantile-regression-starter/notebook
\nhttps://www.kaggle.com/yasufuminakama/osic-lgb-baseline

\n

It would be really helpful to have just a basic explanation of how we predict sigma.

\n

Thank you!

", "rawMarkdown": "Hello,\n\nI've been really struggling to understand how we are supposed to predict sigma. I have a working neural network that predicts FVC just fine but I can't figure out how we predict sigma. I've read a couple posts on in the discussions and some notebooks but the implementations haven't been very clear to me. \nThe best I could surmise was that we need to predict FVC in different quantiles subtract the first and third ones to get some kind of avg. \n\nA couple of the popular ones that people have been passing around on the forums are:\nhttps://www.kaggle.com/ulrich07/osic-multiple-quantile-regression-starter/notebook\nhttps://www.kaggle.com/yasufuminakama/osic-lgb-baseline\n\nIt would be really helpful to have just a basic explanation of how we predict sigma.\n\nThank you!", "votes": null }, { "id": "1026978", "postDate": "09/25/2020 17:52:41", "content": "

There's some good previous discussion threads in the discussion page for the competition that discuss that very question, such as this one: https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/181632#1004612

", "rawMarkdown": "There's some good previous discussion threads in the discussion page for the competition that discuss that very question, such as this one: https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/181632#1004612", "votes": null }, { "id": "1029607", "postDate": "09/27/2020 23:35:21", "content": "

Thank you for posting this. I didn't find that in my earlier search and you laid out a lot of good options

", "rawMarkdown": "Thank you for posting this. I didn't find that in my earlier search and you laid out a lot of good options", "votes": null } ], "comments": [ { "id": 1026978, "author_name": "bjoernholzhauer", "author_url": "", "post_date": "09/25/2020 17:52:41", "content": "

There's some good previous discussion threads in the discussion page for the competition that discuss that very question, such as this one: https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/181632#1004612

", "votes": null, "replies": [ { "id": 1029607, "author_name": "dannyk2018", "author_url": "", "post_date": "09/27/2020 23:35:21", "content": "

Thank you for posting this. I didn't find that in my earlier search and you laid out a lot of good options

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1026092": "Hello,\n\nI've been really struggling to understand how we are supposed to predict sigma. I have a working neural network that predicts FVC just fine but I can't figure out how we predict sigma. I've read a couple posts on in the discussions and some notebooks but the implementations haven't been very clear to me. \nThe best I could surmise was that we need to predict FVC in different quantiles subtract the first and third ones to get some kind of avg. \n\nA couple of the popular ones that people have been passing around on the forums are:\nhttps://www.kaggle.com/ulrich07/osic-multiple-quantile-regression-starter/notebook\nhttps://www.kaggle.com/yasufuminakama/osic-lgb-baseline\n\nIt would be really helpful to have just a basic explanation of how we predict sigma.\n\nThank you!", "1026978": "There's some good previous discussion threads in the discussion page for the competition that discuss that very question, such as this one: https://www.kaggle.com/c/osic-pulmonary-fibrosis-progression/discussion/181632#1004612", "1029607": "Thank you for posting this. I didn't find that in my earlier search and you laid out a lot of good options" }, "source": "meta" }