youth self completion longitudinal weight
Hi US Support,
I have a question about the youth self-completion weight.
My analytical sample comprises of youths and young adults (aged 12-19) who filled in the Self-completion booklet (SC), using data from wave 3 and wave 6 involving the following combinations
A: Wave 3: Youth Sc -> Wave 6: Youth Sc
B: Wave 3:Youth Sc -> Wave 6: Young Adult SC
C: Wave 3:Young Adults SC -> Wave 6: Young Adult Sc
For C, I can use the young adult SC longitudinal weight f_indscub_lw
However, I cannot find a longitudinal weight for youth SC, (Group A), why is that?
Also, what should I do about weighting for youths SC who became young adults in Wave 6 (Group B)?
Thank you in advance.
#2 Updated by Olena Kaminska 7 months ago
Yes, the weight for the situation C is correct.
For the situations A and B we do not have tailored weights to your specific situation. My suggestion is to use the following suboptimal weights:
- Situation A: f_psnenub_lw
- Situation B: f_indscub_lw
Hope this helps,
Dear Olena/Understanding Support staff,
Hope you're all keeping well in these current times!
I have a few questions with regards to this project and weighting if you could be so kind to help.
Firstly, I have weighted as suggested but find that my sample has droped by around 200, is that right? I also get a warning note after my regression that '136 strata omitted because they contain no population members' - should I be concerned?
Secondly, if I'm limiting my sample to only those with records at wave 3 and wave 6 - is it still the longitudinal weight at wave 6 I use when carrying out baseline analyses (at wave 3), and not the cross-sectional weight?
Thirdly, am I right in thinking that the houshold clustering is accounted for via the non-response weighting, therefore I do not need to further account for this when applying complex survey design (where I am already accounting for PSUs within strata)?
Thank you in advance.
STay safe and well!
#6 Updated by Olena Kaminska 6 months ago
- Assignee changed from Linda Ng to Alita Nandi
Thank you for your question.
Yes, the drop in number is correct with weighting - this is because our data is longitudinal (this does not happen if the data is collected at one time point only).
The strata error is a common one - this is because we use very fine strata (which adds precision). There is a very useful guide to how to get over it - I will pass you to Alita who will give you an example of how to deal with it.
And what do you mean by baseline? If you have wave 3 and 6 in your analysis at the same time - use lw weight from wave 6. If you use only wave 3 in your analysis then you can use xw weight from wave 3.
And no, weighting does not correct for clustering. You need to specify psu=w_psu during svyset, as well as weight and strata variables. But you do not need to worry about clustering within households because it is corrected for already through higher level psu's.