topic Re: How to Create replicate weights for BRR method（variance estimation）? in Statistical Procedures

How to Create replicate weights for BRR method（variance estimation）?

keyl — Fri, 08 May 2020 15:45:43 GMT

Recently I need to use BRR（balanced repeated replication）method to estimate the variance of data from a stratified sample design research(NHANES) .

I have got some examples,and i find that replicate weights have already been given (16 replicates) for each observation.

But the replicate weights are not included in primary data absolutely. I tried the brr method using the example's dataset with the following code.

proc surveymeans data =TMP1.DEMOADV varmethod = brr
(outweight = RepWeights);
weight w0304_0;
strata SDMVSTRA;
cluster SDMVPSU;
var SEQN;
run;

The new replicate weights(RepWt) generated from this code are quite different from the weights given in the example.（only 2 values——0 and twice the primary weight）

How can I supposed to get a result like the examples(w0304) for the next steps needing replicates？（16 replicate weights are not equal to the others）

Re: How to Create replicate weights for BRR method（variance estimation）?

TonyAn — Fri, 08 May 2020 17:30:36 GMT

The SURVEYMEANS Procedure

REPWEIGHTS Statement

REPWEIGHTS variables </ options>;

Re: How to Create replicate weights for BRR method（variance estimation）?

keyl — Sun, 10 May 2020 04:56:48 GMT

Would you please show a exact sample code？

The result was not different after I added repweights statement. As far as I am concerned, the repweights statement is used when the replicate weights are already given,so you don't have to offer STRATA and CLUSTER.

And I just want to know how can I generate all these replicate weights.

Thanks for your reply.

Re: How to Create replicate weights for BRR method（variance estimation）?

TonyAn — Mon, 11 May 2020 15:45:34 GMT

If the agency already provided you the replicate weights, why do you want to recreate them?

Each sets of replicate weights can be different for BRR and Bootstrap methods. If you want to generate the exact set of replicate weights as published in the data set, you have to contact the publisher of those replicate weights.