Properly Specifying BRR Replicate Weights in R Survey Package

Question

I am using Lumley's text on the Survey Package; I'm not sure what to indicate for the combined.weights argument. I looked in the help window on it and it said the argument is true if the BRR weights include the sampling weight (and it also said that it is usually the case). I am using the National Center for Education Statistics ELS 2002 Dataset. Any thoughts on the proper combined weights argument? Also, if the combined.weights argument is true, then scale and rscales have to be specified according to the R help page and I really am not sure how to go about doing that either. Was not able to find anything very detailed in the help window and other examples I have found vary greatly depending on the dataset involved. There was another question posted on specifying BRR weights on this site; the person answering showed specifying the BRR weights with the combined.weights argument=TRUE, but then did not specify scale and rscales. Any guidance is greatly appreciated; I really want my analyses to be accurate, so these specifications about the dataset are critical. Thank you!

Answer 1

I was able to obtain this answer through the U. of Michigan Survey Methodology Program: combined.weights=TRUE is the default and is probably the right setting for this dataset. If the weights in a replicate sum to something that looks like the pop total of students (or whatever the units are), then combined.weights=TRUE is correct.

According to Lumley's book, sec, 2.3.1 (link below), R survey will figure out what scale and rscale for BRR and Fay. For fay she does have to specify rho, the Fay factor.

https://books.google.com/books?id=L96ludyhFBsC&pg=PP35&lpg=PP35&dq=combined.weights+in+R+survey&source=bl&ots=ic-u-jnCv1&sig=g9RfrFezc_42Qr-__U3d5QTMER4&hl=en&sa=X&ved=0ahUKEwjJ-o6wyaDMAhUKdT4KHUmgBK4Q6AEIWjAJ#v=onepage&q=combined.weights%20in%20R%20survey&f=false