Support #1159

Weights for cross-sectional and longitudinal analyses

Added by Luca Bernardi about 5 years ago. Updated about 5 years ago.

Start date:
% Done:



Dear Understanding Society Support Team,

I am using data from adult main interviews from all waves. I am estimating the effect of depression on party identification. I am analyzing the data both cross-sectionally and longitudinally. However, I am unsure about which weight(s) to use, also given the low number of clinically depressed individuals. Is it correct that in both cases, since I am using data from more than one wave, I should use a longitudinal weight? Also, by reading the User Guide, in Wave 6 there is a change in the definition of the cross-sectional population represented. If this somehow complicates the issue, I have no problem with analyzing data only from Wave 1 to Wave 5. Could you please give me some recommendations?

Many thanks and best wishes,


Updated by Olena Kaminska about 5 years ago


Thank you for your question.
If you are interested in estimates in a particular time point - you should use cross-sectional weights. In other words if you estimate the same model for each wave separately and you include only information from that particular wave in the model - you should use the cross-sectional weight from that wave.
If you are using information from across a few waves - you should use a longitudinal weight from the last wave in your analysis.
The change in definition at wave 6 relates to immigrants: from wave 6 'ui' weights include immigrants since 2009 but 'ub' weights don't include them. This is only relevant if you put together many cross-sectional estimates and compare them over time (e.g. looking at a trend). This does not influence longitudinal analysis. If you want to be sure to have comparable cross-sectional estimates (i.e. those that exclude immigrants since 2009) then you can use 'ub' weights throughout.

Hope this helps,


Updated by Luca Bernardi about 5 years ago

Dear Olena,

Many thanks for replying so quickly. That's very helpful. If I understand correctly, since I am using information from across a few waves, I can apply the "indinub_lw" weight. Is this correct?

Thanks and best wishes,


Updated by Olena Kaminska about 5 years ago


Yes, the weight is correct, and it should come from the last wave that you use in your analysis. Note, 'ub_lw' weight is available for analysis from wave 2 onwards.



Updated by Luca Bernardi about 5 years ago

Dear Olena,

Sounds good. Many thanks again.



Updated by Stephanie Auty about 5 years ago

  • Status changed from New to Resolved
  • % Done changed from 0 to 100
  • Private changed from Yes to No

Also available in: Atom PDF