Understanding Society User Support

Thanks for your question. Yes, we do not provide separate cross-sectional weights for separate samples simply because our 'ub' weight that combines BHPS, GPS and EMB represents the population and has an advantage of larger sample size. Thus we suggest you use 'ub' equivalent cross-sectional weight. Please note that if you pool information from different waves and use it in one analysis you may want to scale the weights.
Alternatively you can use the equivalent longitudinal weight for 2001 BHPS sample. Longitudinal weights also represent cross-sectional population (although this would exclude immigrants to the UK since 2001).
Hope this helps,
Olena

For how to scale the 'ub' weight please see our reply to issue #228.
Olena

Hi Olena,

Thank you very much for the quick response.

However we are still a little unsure how to proceed. We only want to include BHPS respondents in our analysis because (a) we are also using panel surveys from other countries which follow a single sample from 2001 onwards and (b) we do not want to pick up effects caused by recent immigration. For 2001-2008 we understand that we can just use the XRWTUK1 variable in BHPS. However after the BHPS participants we are tracking enter UKHLS in W2 (2010) we don't know which weight would be best. Is it appropriate to just use the indinub_xw variable with only BHPS individuals-not GPS, EMB or GPC- in the sample?

We understand that we could use the relevant longitudinal weight but this would result in the loss of lots of cases with incomplete response patterns.

Thanks!

Olena Kaminska wrote:

Thanks for your question. Yes, we do not provide separate cross-sectional weights for separate samples simply because our 'ub' weight that combines BHPS, GPS and EMB represents the population and has an advantage of larger sample size. Thus we suggest you use 'ub' equivalent cross-sectional weight. Please note that if you pool information from different waves and use it in one analysis you may want to scale the weights.
Alternatively you can use the equivalent longitudinal weight for 2001 BHPS sample. Longitudinal weights also represent cross-sectional population (although this would exclude immigrants to the UK since 2001).
Hope this helps,
Olena

I suggest you use W_indin01_lw. This is a longitudinal weight but it also represents the cross-sectional population. It is equivalent to LRWTUK1 in BHPS.
Hope this helps,
Olena