Support #1417

Time-varying predictor variables at different waves to outcome

Added by George Comp about 3 years ago. Updated almost 2 years ago.

Start date:
% Done:



I am using longitudinal models and have a specific question regarding Understanding Society data and a specific way I would like to use variables.

I am looking at the effects of screen usage on wellbeing scores in youth but then some of them move into the main adult survey when they turn 16.

There is a variable called subjective wellbeing which is used in both the youth and the adult survey.

The subjective wellbeing questions are asked at waves 2, 4, 6 and 8 which is useful for me and it means that when individuals move from the youth to the adult survey, they are asked the same questions and it could be a good outcome to use.

My question is- a lot of my predictor variables of interest such as “games console usage”_ypconstm_ gets asked at waves 1,3,5, and 7. And I may potentially use games console usage as a time-varying variable. This is also the case for computer screen usage ypcpgs which gets asked at waves 1,3,5,7. Let’s say I wanted to use one of these as a time-varying variable of interest.

But is it an acceptable practice where some of my preferred predictors don’t occur at the wave of the outcome which is at 2,4,6,8, to then use them in the same model as time varying, when they occur at 1,3,5,7?

Is this normal practice? or is there a name for it? Also, are there any studies that do this which might clarify this question?

Any help would be much appreciated and thank you.


Updated by Alita Nandi about 3 years ago

  • Status changed from New to Feedback
  • Assignee set to George Comp
  • % Done changed from 0 to 50
  • Private changed from Yes to No

Dear Goerge,

Our remit is to answer queries related to Understanding Society data and provide general advice about how to manage the data (and about weights provided with the survey). Given the number of users we have we do not have resources to advise on individual users' analysis or software use specifically. You can sign up to our JISCmail which we have set up for researchers to ask and answer such questions to/from each other. If you want to join that please email

Sorry, there is no definitive answer to your question as it depends on your research question, the existing evidence on the patterns of behaviour as measured by these variables which will provide guidance on the types of assumption you can make. For example, if you think any of these variables is not likely to change over a 2 year period then you can treat the value at t-1 = value at t. Also, a general rule is that questions that are expected to change very often are asked every year, others less frequently.

Best wishes,


Updated by Alita Nandi almost 3 years ago

  • % Done changed from 50 to 90

Updated by Understanding Society User Support Team almost 2 years ago

  • Status changed from Feedback to Resolved
  • Assignee deleted (George Comp)
  • % Done changed from 90 to 100

Also available in: Atom PDF