Understanding Society User Support

Arpita,

Thank you for your question. 'ui' weight is better in two ways: it has higher sample size (so higher statistical power), and also represents UK population better (from wave 6 onwards) as it includes more recent immigrants. There is no advantage in using 'ub'.
If you are able to merge each of our participants to external data you should use our weights to represent the population. If you have lots of missingess in your independent variable after merging, then I would suggest to make an extra adjustment for this missingess (either through imputation or additional weighting). If you are doing this adjustment through weighting the new weighting adjustment will have to be multiplied by our weight to get the weight for your analysis.

Hope this helps,
Olena

Dear Olena,

Thank you very much for your reply. It helps a lot.

Best, Arpita

Olena Kaminska wrote:

Arpita,

Thank you for your question. 'ui' weight is better in two ways: it has higher sample size (so higher statistical power), and also represents UK population better (from wave 6 onwards) as it includes more recent immigrants. There is no advantage in using 'ub'.
If you are able to merge each of our participants to external data you should use our weights to represent the population. If you have lots of missingess in your independent variable after merging, then I would suggest to make an extra adjustment for this missingess (either through imputation or additional weighting). If you are doing this adjustment through weighting the new weighting adjustment will have to be multiplied by our weight to get the weight for your analysis.

Hope this helps,
Olena

Arpita Ghosh wrote:

Dear User Support Team,

I hope you are doing well in these difficult times. My questions are related to this old one (https://iserswww.essex.ac.uk/support/issues/1257) but extend a little further. I am merging household and individual response files for each of the waves of 1-9 of UKHLS and then appending them together. Following the discussion in the old thread above, I understand that I should use the cross sectional adult main survey weights for each wave in my appended UKHLS files, i.e., indinus_xw for waves 1 and 2, indinub_xw for waves 3, 4, 5 and indinui_xw for waves 6, 7, 8, 9. My questions are:
1. What is the repercussion of using indinub_xw for waves 6-9 instead of indinui_xw weight, as I am concentrating on sf12mcs_dv, sf12pcs_dv etc variables which have been collected throughout all the waves?
2. I am merging the appended UKHLS file to another data set which has the independent variable that I want to use in regressions with sf12mcs_dv (for e.g.) being the dependent variable. In this case, am I right to understand that I should create a new weight (which will correspond to both the dependent and independent variables) in the final data in order to analyse?

Thank you very much for your time and consideration and I will look forward to hear from you.

Best, Arpita