Support #137


Added by Jennifer Brown over 10 years ago. Updated over 10 years ago.

Data analysis
Start date:
% Done:




I'm working with b_indinus_xw as I'm interested in variables from the adult main interview. I'm also using as a filter b_ivfio=1 for full interviews. When carrying out analyses I get a warning message: warning #3211 which refers to the value of the weight being zero, negative or missing in some cases. Also comparing the results with and without weights gives a big difference: 50,389 valid cases against 38,788.
Reading the user guide and previous issues in the user forum I understand that in wave 1 a weight of 0 is given to temporary sample members who will not be followed in the future. Is this the case for wave 2 too? is this the reason for my warning message and the discrepancies in valid cases between results with weight and without weight? and am I justified in using this weight? I'm looking at sports and leisure activities amongst adults.
Thanks, Jen


Updated by Olena Kaminska over 10 years ago


Thanks for the question. As long as you want to infer to the population you must use weights. The value of 0 is correct, but calculated in a very complex way - the full description is in the technical part of the documentation on weighting. But for your analysis you don't need to know the details.

Just in short: _xw is for cross-sectional analysis. For this weight most TSMs have non-zero values, but 0 values are mostly present for BHPS part of the sample, and for white British selected through ethnic boost. We are currently working on the weight which will have fewer 0's (by combining BHPS with UKHLS samples), but because of the theoretical and practical complexities of the procedure this is not ready yet.

In short, ignore the warning message, and continue using weights.



Updated by Redmine Admin over 10 years ago

  • % Done changed from 0 to 50

Updated by Redmine Admin over 10 years ago

  • Status changed from New to Closed
  • % Done changed from 50 to 100

Also available in: Atom PDF