Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- How to obtain c-statistics from multiple 'proc log...

- RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

04-19-2018 12:54 PM

Hi SAS communities,

I have the following challenge:

I have a logistical regression model including a binary outcome and some categorical and continuous covariates.

The logistical regression model achieved a c-statistic (or area under the ROC curve) of 0.94.

I have then used bootstrapping (unrestricted random sampling) to create 50 data sets (later I will do 1000) put together in one data set, sorted by replicate number (1-50).

Then I run the proc logistic on the large data set, using a 'by replicate' statement.

In the SAS results window I how have 50 separate c-statistics.

Which leads me to the question:

How do I combine the 50 separate c-statistics in the results window, so that I can obtain a single median c-statistics for all the bootstrapped data sets? Preferably also with confidence limits...

I have attached the SAS code in a word file.

Best,

Sebastian Wiberg, MD

Accepted Solutions

Solution

04-19-2018
02:37 PM

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to scwiberg

04-19-2018 12:59 PM

How to store your results into a table:

https://blogs.sas.com/content/iml/2017/01/09/ods-output-any-statistic.html

Look at the PERSIST option on the ODS OUTPUT statement that will allow you to stack them all into one until it's done.

@scwiberg wrote:

I have the following challenge:

I have a logistical regression model including a binary outcome and some categorical and continuous covariates.

The logistical regression model achieved a c-statistic (or area under the ROC curve) of 0.94.

I have then used bootstrapping (unrestricted random sampling) to create 50 data sets (later I will do 1000) put together in one data set, sorted by replicate number (1-50).

Then I run the proc logistic on the large data set, using a 'by replicate' statement.

In the SAS results window I how have 50 separate c-statistics.

Which leads me to the question:

How do I combine the 50 separate c-statistics in the results window, so that I can obtain a single median c-statistics for all the bootstrapped data sets? Preferably also with confidence limits...

I have attached the SAS code in a word file.

Best,

Sebastian Wiberg, MD

All Replies

Solution

04-19-2018
02:37 PM

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to scwiberg

04-19-2018 12:59 PM

How to store your results into a table:

https://blogs.sas.com/content/iml/2017/01/09/ods-output-any-statistic.html

Look at the PERSIST option on the ODS OUTPUT statement that will allow you to stack them all into one until it's done.

@scwiberg wrote:

I have the following challenge:

I have a logistical regression model including a binary outcome and some categorical and continuous covariates.

The logistical regression model achieved a c-statistic (or area under the ROC curve) of 0.94.

I have then used bootstrapping (unrestricted random sampling) to create 50 data sets (later I will do 1000) put together in one data set, sorted by replicate number (1-50).

Then I run the proc logistic on the large data set, using a 'by replicate' statement.

In the SAS results window I how have 50 separate c-statistics.

Which leads me to the question:

How do I combine the 50 separate c-statistics in the results window, so that I can obtain a single median c-statistics for all the bootstrapped data sets? Preferably also with confidence limits...

I have attached the SAS code in a word file.

Best,

Sebastian Wiberg, MD

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Reeza

04-19-2018 02:37 PM

Thank you very much! That did the trick...

/Sebastian