I am using Stata 13.1 (for Windows) to perform secondary data analysis of the Fragile Families and Child Wellbeing Study dataset by Princeton University and Columbia University (http://www.fragilefamilies.princeton.edu/index.asp) . This dataset has five waves of data with a sample size of 4989. The data files include national level survey weights (both basic and replicate weights) at each of the five waves.
Within their documentation is a guide “Using the Fragile Families Weights” (http://www.fragilefamilies.princeton...ts_rev0709.pdf) which details the following command:
I did try running the mixed command using the level one and level two weights as basic and replicate weights, respectively, using the following command:
I would appreciate any thoughts or examples on how to conduct such an analysis! Thank you.
Within their documentation is a guide “Using the Fragile Families Weights” (http://www.fragilefamilies.princeton...ts_rev0709.pdf) which details the following command:
svyset [pweight=BASICWEIGHT], jkrw(REPLCATEWEIGHT, multiplier(1)) vce(jack) mseThe analysis I am trying to perform with the Fragile Families dataset uses multilevel modeling with the mixed command, and I understand that the svy command cannot be used with mixed. As a result, I am following along with the instructions provided in Stata’s manual (http://www.stata.com/manuals13/memix...plesSurveydata) which through the text and example 12 recommends specifying level 1 and level 2 weights with the following example code:
mixed isei female high_school college one_for both_for test_lang [pw=w_fstuwt] || id_school:, pweight(wnrschbw) pwscale(size)
where: pw= the level one weight
pweight(level two weight)My primary question is how do I reconcile the two approaches to weights? Would the basic weight be correct as the level one weight and the replicate weight work as the level two?
I did try running the mixed command using the level one and level two weights as basic and replicate weights, respectively, using the following command:
mixed intlze c.cagem##c.mbltobexp [pw=m4natwt] || idnum: cagem , cov(un) pweight(m4natwt_rep1) pwscale(size)However, Stata returned an error “pweights must be positive.” I used summarize to check both weights used and the ranges for both are 0 to 8057.84 and 1.71 to 8005.14. More importantly, however, I want to make sure this approach is correct.
I would appreciate any thoughts or examples on how to conduct such an analysis! Thank you.
