svyset Command

[Pat Burns]

Can anyone tell me the difference between the following two svyset commands? I am not sure which one to use with my research using the NCES sponsored ECLS-K: 2011 dataset.

svyset_n[pweight=(W1P0)], brrweight(W1P1 - W1P80) vce(jk2) mse
-OR-
svyset_n[pweight=(W1P0)], brrweight(W1P1 - W1P80) vce(jk2) singleunit(centered)

[David Radwin]

It looks like neither command is correct. Both seem to combine BRR (balanced repeated replication) estimation with jackknife estimation. And jackknife estimation in Stata is indicated by the option vce(jackknife), not vce(jk2).

I suggest you find the Stata syntax in the documentation or data files at https://nces.ed.gov/ecls/dataproducts.asp, then compare several standard errors you created in Stata with those in a published report to make sure they match. For the recent NCES postsecondary sample studies, by way of comparison, you must use BRR estimation and specify the mse option for the standard errors to match published standard errors (although there are other estimation options available). (NCES = the National Center for Education Statistics in the U.S. Department of Education.)

If you cannot find the Stata syntax, I recommend you contact NCES directly using the contact information on that page.

[Pat Burns]

Thank you!

[Pat Burns]

This is the documentation I found in the users manual.... I am still unsure then what state command to use? Any thoughts?

Thanks in advance

4.2.4.3 Specifications for Computing Standard Errors

For the jackknife replication method, the full sample weight, the replicate weights, and the method of replication must be specified. All analyses of the ECLS-K:2011 data using the replication method should be done using JK2. As an example, to compute child-level estimates (e.g., mean reading scores) and their standard errors, users need to specify W1C0 as the full sample weight, W1C1 to W1C80 as the replicate weights, and JK2 as the method of replication.

[Marcos Almeida]

About JK2, there is an old thread on this , then on the same verge as #2.

[Pat Burns]

Thank you... I will contact NCES. Appreciate everyone's help!