- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Posted 07-08-2009 05:05 PM
(1432 views)
Hi ,
I am using proc logistics with weight statement to take care of over sampling in the sample dataset.. I am using score statement in the same procedure to score the entire population dataset.
My question is during scoring will the wieghts be applied to entire population.
If it does then how can i prevent it as I am using the entire population.
Thanks,
Amit
I am using proc logistics with weight statement to take care of over sampling in the sample dataset.. I am using score statement in the same procedure to score the entire population dataset.
My question is during scoring will the wieghts be applied to entire population.
If it does then how can i prevent it as I am using the entire population.
Thanks,
Amit
1 REPLY 1
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
This is discussed in the description of the WEIGHT statement and the "Details: Scoring Data Sets" section in the LOGISTIC documentation:
http://support.sas.com/onlinedoc/913/getDoc/en/statug.hlp/logistic_index.htm
Computation of predicted values (scoring) done by the SCORE statement does not use weights. However, weights, if available, are used to compute the fit statistics and ROC curve that the SCORE statement can also produce.
If you specify the MODEL, SCORE, and WEIGHT statements together to fit the model and score data in one step, then the weights affect only the model estimation and therefore the parameter estimates of the model. They are not used again in the scoring process.
See this usage note for more on oversampling: http://support.sas.com/kb/22/601.html
http://support.sas.com/onlinedoc/913/getDoc/en/statug.hlp/logistic_index.htm
Computation of predicted values (scoring) done by the SCORE statement does not use weights. However, weights, if available, are used to compute the fit statistics and ROC curve that the SCORE statement can also produce.
If you specify the MODEL, SCORE, and WEIGHT statements together to fit the model and score data in one step, then the weights affect only the model estimation and therefore the parameter estimates of the model. They are not used again in the scoring process.
See this usage note for more on oversampling: http://support.sas.com/kb/22/601.html