Individual and Household Weights for combining EMBS plus ethnic minority individuals in GPS & by 5 ethnic groups
I am interested in conducting 3 analyses:
1) an analysis that compares all UK ethnic respondents (both from the EMBS and within GPS) compared to non-minority white respondents in the GPS
2) same as above, but comparing each of the 5 ethnic groups in the UK (both from the EMBS and within GPS) compared to non-minority white respondents in the GPS
3) comparing across each of the 5 ethnic groups from the EMBS only
In the working paper regarding weighting strategy for the study (pages 9-10), you mention providing many different weights. I was wondering if you could tell me which weighting variables I could use for my analysis above? In other words, I'm unsure of the weights to use in analyzing:
1) The EMBS plus ethnic minority individuals in the UKHLS-GPS in areas of low ethnic minority density, i.e. the areas excluded from the boost sample (to represent the total UK ethnic minority population);
2) Each of the five main ethnic groups separately from the combined EMBS plus ethnic minority individuals in the UKHLS-GPS
3) Each of the five main ethnic groups separately from the EMBS
at both a) the individual level and b) household level?
Many thanks for your assistance.
#1 Updated by Redmine Admin almost 8 years ago
- Assignee set to Redmine Admin
- % Done changed from 0 to 80
Thank you for your request.
You should first read the section on "weighting adjustments" (starting p.14) in the user guide (https://www.understandingsociety.ac.uk/files/data/documentation/wave1/User_manual_Understanding_Society_Wave_1.pdf). This describes the weight variables available on the wave 1 data set.
You should use the same weights for each of your two types of analysis. Each weight is designed to give correct representation of the total population as well as all subsets within the population (such as all non-whites, each of 5 ethnic groups, etc).
The correct weight to use depends on whether or not your analysis includes variable(s) that are collected as part of the "extra 5 minutes" questions (asked only of the ethnic minority boost sample, ethnic minorities in the 'low EM density' stratum of the main sample, and the general population comparison sample). If you are using such variables, then for individual-level analysis you should use a_ind5mus_xw. If not, use a_indinus_xw (or, if your variables are restricted to those that are available also in the proxy interview, a_indpxus_xw, as this will add in proxy respondents to the analysis). For household-level analysis, use a_hhdenus_xw.
#2 Updated by Sung Park almost 8 years ago
Thank you for your response. I just had one follow-up question: How do I identify which exact questions or entire modules are specifically classified as those asked within the "extra 5 minutes"? I haven't been able to find a list contained within the documentation of this.
#3 Updated by Peter Lynn almost 8 years ago
I think that you can only do this through the variable-level online documentation. For each variable, check whether the universe is described as "if ( EMBoost = 1 | GPCompare = 1 | (LDA = 1 & RACEL > 4 & RACEL < 98) ).
For example, if you look at the Discrimination or Remittances modules (at https://www.understandingsociety.ac.uk/documentation/mainstage/dataset-documentation) you will see that all variables are of this type. In the Migration History module, the universe is defined using different variables, but I think it amounts to the same set of respondents. HTH.