Calcite | Level 5

## Creating an ROC Curve with Weighted Sample

I am trying to generate an ROC curve using the PROC LOGISTIC procedure (see example below that I am using as a guide):

descending data = pinedat plots=effect plots=ROC(id=prob);

class alc /descending param = glm;

model y = age alc ;

roc 'Age' age;

roc 'Alcohol' alc;

roc 'Age & Alcohol' age alc ;

roccontrast reference('Age & Alcohol') /estimate

However, my sample is weighted, and I'm not sure how to modify this to be used with PROC SURVEYLOGISTIC. Any help would be much appreciated. Thanks!

1 ACCEPTED SOLUTION

Accepted Solutions
Super User

## Re: Creating an ROC Curve with Weighted Sample

@gabybarber wrote:
My data (PRAMS data from the CDC) does have a complex sample design. Do you happen to know how to do an ROC curve with data that has a complex sample design?

Thanks!

I think this paper may help:  https://support.sas.com/resources/papers/proceedings14/SAS404-2014.pdf

Google: SAS Logistic ROC Pred Surveylogistic

4 REPLIES 4
Super User

## Re: Creating an ROC Curve with Weighted Sample

Does your data come from a complex weighting scheme? If not then perhaps all you need is the Weight statement in Proc Logistic.

Surveylogistic doesn't currently do plots.

https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=statug&docsetTarget=statu...

Calcite | Level 5

## Re: Creating an ROC Curve with Weighted Sample

My data (PRAMS data from the CDC) does have a complex sample design. Do you happen to know how to do an ROC curve with data that has a complex sample design?

Thanks!
Super User

## Re: Creating an ROC Curve with Weighted Sample

@gabybarber wrote:
My data (PRAMS data from the CDC) does have a complex sample design. Do you happen to know how to do an ROC curve with data that has a complex sample design?

Thanks!

I think this paper may help:  https://support.sas.com/resources/papers/proceedings14/SAS404-2014.pdf

Google: SAS Logistic ROC Pred Surveylogistic

Calcite | Level 5

## Re: Creating an ROC Curve with Weighted Sample

Thanks!
Discussion stats
• 4 replies
• 709 views
• 0 likes
• 2 in conversation