Using the EMB sample independently
Hello, I have a question regarding the EMB sample I am hoping you can help me with? I was hoping to conduct an analysis of the EMB sample using SEM in order to look at the role of racism (using discrimination and harassment modules) as a mechanism affecting mental health (specifically using wave 3 data). However, I have seen in the main stage user guide that the EMB sample should not be analysed independently and that no weight is provided for this. However, a weight is provided for use of the extra five minutes questions (n_ind5mus_xw). I have also seen two papers which as far as I can tell seem to use the EMB sample independently (https://www-ncbi-nlm-nih-gov.sheffield.idm.oclc.org/pmc/articles/PMC4984732/pdf/AJPH.2016.303121.pdf and https://www.tandfonline.com/doi/pdf/10.1080/1369183X.2018.1539287?needAccess=true).
I was wondering if you could help me to understand if it is actually possible to perform an analysis only on the boost sample as the two papers mentioned have or if I am misunderstanding something?
#1 Updated by Alita Nandi 3 months ago
- Private changed from Yes to No
- % Done changed from 0 to 50
- Assignee set to Natalie Bennett
- Status changed from New to Feedback
You are correct that EMB sample should not be analysed by itself. The JEMS paper you mentioned does not do that, I could not access the second paper and so I cannot say what they did. If you analyse the EMB sample by itself, the estimates based on that sample will have coverage error as the EMB sample by design does not include ethnic minorities living in low ethnic minority concentration areas. Additionally, as the weights provided rquire you to include GPS in the analysis, you will not have access to appropriate weights.
The extra five minutes weights requires you to use the extra five minutes sample = EMBS + GPC (500 randomly selected hh from GPS) + ethnic minorities in GPS living in low ethnic minority concentration areas.
Here is our guidance on weighting:
and on using Ethnicity and Immgiration data in the study:
You may also be interested in papers that have looked at the association of mental health aspects and experiences of harassment and discrimination using our data:
https://www.iser.essex.ac.uk/research/publications/working-papers/iser/2016-14 (currently submitted to a peer-reviewd journal)
If you have additional questions please let us know.
#2 Updated by Natalie Bennett 3 months ago
Thanks so much for your reply. Apologies, I think I am confusing the extra 5 minutes sample and the EMB sample. If I want to analyse only those who have been asked the extra five minutes questions (so EMB + 500 random GPC + GPS in low minority density areas), is this possible so long as the weight n_ind5mus_xw is used?
Thanks also for the paper suggestions, they are very helpful.
#3 Updated by Alita Nandi 3 months ago
- % Done changed from 50 to 80
"If I want to analyse only those who have been asked the extra five minutes questions (so EMB + 500 random GPC + GPS in low minority density areas), is this possible so long as the weight n_ind5mus_xw is used?"
Yes, that is correct.
#4 Updated by Natalie Bennett 3 months ago
That's great, thanks so much for your help. One last question; if I wanted to stratify my analysis by ethnicity and produce regression models separately for certain ethnic groups, would this still be possible with the extra five minutes sample (if I include members of the ethnic group in question from both the EMB and the GPS in low density areas)?
#5 Updated by Alita Nandi 3 months ago
" if I wanted to stratify my analysis by ethnicity and produce regression models separately for certain ethnic groups, would this still be possible with the extra five minutes sample (if I include members of the ethnic group in question from both the EMB and the GPS in low density areas)?"
Yes, that is fine.