topic Re: Proc Survey Select - Stratified Random Sampling in New SAS User

Proc Survey Select - Stratified Random Sampling

tlnarayana26 — Tue, 21 May 2019 12:49:17 GMT

Hi to all,

Sas code

/*STRATA METHOD */
PROC SORT DATA=PROD_DATA OUT=PROD_DATA_SORT;
BY PROD_FLAG;
RUN;

PROC SURVEYSELECT DATA=PROD_DATA_SORT OUT=STRATA_SPLIT SAMPRATE=.7 SEED=34258 OUTALL;
STRATA PROD_FLAG;
RUN;

DATA TRAINING VALIDATION;
SET STRATA_SPLIT;
IF SELECTED = 1 THEN OUTPUT TRAINING;
ELSE OUTPUT VALIDATION;
RUN;

I have one data set it contains 100 observations. Spiting data into 70% and 30% using PROC SURVEYSELECT method with Strata statement.

Then data is split into 70 and 30 .

But i have two problems

1.Run same code into another system/ laptop. the data split is 70 and 30.

But In both of 70 agents are differed.

How to get same agents on all laptops?

2.Adding some variables to data set and spit is 70 and 30

But In both of 70 agents are differed.

How to get same agents on all laptops?

please give your valuable suggestions

Thanks,

Lakshminrayana.

Re: Proc Survey Select - Stratified Random Sampling

ballardw — Tue, 21 May 2019 14:15:52 GMT

If exact same data is needed on multiple machines the best approach is likely to generate one data set and copy to other computers. Possibly that data set should only contain 'agents' what ever those may be.

Are you specifying the exact same SEED= value on the other machines? SEED may not work on different machines for a number of reasons. ANY difference in the input data set will change the output. Sorting by one variable is not going to guarantee that the order of the data is exactly the same if there are any other variables in the data.

I do not understand how your topic 2 differs from topic 1.

Re: Proc Survey Select - Stratified Random Sampling

tlnarayana26 — Fri, 24 May 2019 06:46:41 GMT

Thanks,
ballardw