07-08-2014 01:01 PM
Hi, I have two questions about surveylogistic procedure and weight statement.
First, I am doing analysis national data by using logistic regression(binary dependent).
I want to use 'sampling weight' but my data sets don't have 'strata' and 'cluster' variabels.
In my situation, what should I do? should I use proc logistic procedure? Or can I use 'proc surveylogistic' procedure with weight variable(but without strata and(or) cluster variables)?
Second, in 'proc logistic' procedure, what does 'weight' statement mean? Does 'weight' mean 'frequency weight'?
I really wonder about weight statement's meaning in SAS.
And I really hope your help and opinion!! Thank you!
07-08-2014 01:30 PM
Proc logistic weights each obs in input data by the value of weight variable. When you a binary variable in your data set, you have actually two strata (1 and 0). So you can apply weights to these strata based on weight variable calculated for class 1 and class 0.
07-08-2014 02:28 PM
Thanks for your reply.
If I unerstand correlty(I hope so), Does your answer mean that I can use 'proc surveylogistic' with weight variable(but without strata and cluster variables)?
07-08-2014 02:42 PM
Sampling weights are used to calibrate parameter estimates calculated based on a sample. Just check proportion of 1's and 0's in your population data and compare them with sample proportions. This will provide a baseline to create weight variable that can be used in proc logistic. Hope this helps,
07-17-2015 12:38 PM
Thank you Reeza! I didn;t want to spend all of the time and effort to download SAS UE if it didn't have these functions. Hopefully the programs I wrote and ran on SAS 9.2 will run smoothly in SAS UE. Any tips you have on transitioning to SAS UE from 9.2?