07-20-2017 10:34 PM - edited 07-20-2017 10:35 PM
Using longtidunal monthly sample data. I can get the coefficient for total frequency in Month1 and Month2, but the change of the frequency between the two months has high correlation and therefore the variance of the CHANGE is not equal to the upper and lower bounds of the monthly estimates.
How can I calculate the variance of the month-to-month change using bootstraps and proc surveyfreq or surveymeans?
07-21-2017 11:46 AM
Are you using variance to mean something other than the normal statistical definition of variance as sigma-squared?
This almost sounds more like a time-series question. If so then the survey procs can only possibly be a part of the solution.
You may have to provide a little more concrete example, such as some data and the desired result along with what code you have been running to illustrate your situation.