Project

General

Profile

Support #419

Adult self completion wave 2 and weights

Added by afshin zilanawala over 8 years ago. Updated over 8 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Category:
Weights
Start date:
09/16/2015
% Done:

100%


Description

hi user support,

1. i am doing a cross-sectional analysis using the wave 2 adult self completion (SC) interview. what weight do i use? theoretically there should be a b_indscub_xw but only b_indscus_xw is available. if i use the latter to weight my means, then i lose 16,000 observations, which i'm assuming are the bhps people?

2. if i combine questions from the wave 2 adult SC interview with the biomarker data, then i should be using indbdub_xw? what sample am i representing when i use this weight?

thanks for all the help!

#1

Updated by Olena Kaminska over 8 years ago

1. 16,000 that you mentioned include 4000 proxies (that didn't get self-completion) + nonrespondents to self-completion questionnaire + those from BHPS. But yes, the self-completion cross-sectional weight is available only for UKHLS and not for BHPS+UKHLS samples. This is because BHPS has never had self-completion weight and to create a combined weight we would need to create such weight going back to 1991 (i.e. a long and complicated process). Theoretically for most analysis (unless you are looking at tiny subgroups) UKHLS should be sufficiently large. It is also 'better quality' in terms of overall lower nonresponse and more up-to-date in terms of immigrants - so has some advantages.

2. Yes, when using SC questionnaire and biomarker data you don't have an optimal weight. There are two suboptimal options for you:
a) you can use sc weight (b_indscub_xw) (and yes, you loose BHPS people here) - this corrects for all nonresponse except for the nonresponse to biomarkers conditional on self-completion in w2;
b) you can use indbdub_xw (for blood) or indnsub_xw (for nurse visit) - this will correct for full nonresponse except for nonresponse to self-completion conditional on nurse visit / giving blood.

You can choose between the two options above simply based on the number of cases; or you can choose based on more important variables for you (either from SC or nurse visit/blood).

We apologize that we do not produce weights for any combination of questionnaires and instruments potentially of interest. This is mainly because doing so will generate easily over 1,000 weights and more over time.

#2

Updated by afshin zilanawala over 8 years ago

thank you so much for your detailed response. very helpful! i have one follow up question w/r to 2a:

2. Yes, when using SC questionnaire and biomarker data you don't have an optimal weight. There are two suboptimal options for you:
a) you can use sc weight (b_indscub_xw) (and yes, you loose BHPS people here) - this corrects for all nonresponse except for the nonresponse to biomarkers conditional on self-completion in w2;

if i want to correct for nonresponse except for nonresponse to biomarkers conditional on SC at W2, then i should use b_indscus_xw NOT b_indscub_xw as you indicated, correct?

thank you!

#3

Updated by Olena Kaminska over 8 years ago

Yes, you are right. You should use b_indscus_lw. Thanks for pointing it.

#4

Updated by Redmine Admin over 8 years ago

  • Status changed from New to Closed
  • % Done changed from 0 to 100
#5

Updated by Gundi Knies over 8 years ago

  • Target version set to M2

Also available in: Atom PDF