Understanding Society User Support

Georgina,

Thank you for your question. If you use information from a later adult interview, it would be best to use the weight from the last wave of full adult interview in your analysis. Just a note, that it would be easiest to use this weight for everyone in you analysis (so the weight from the same wave). If you are concerned about numbers other options are possible, and they are similar to pooling, but pooling needs to be done based on population characterstic, and not on the rule of 'last interview that a person participated'.

Hope this helps,
Olena

Hi Olena,

Thank you for this. I don't think I explained my research design very well so I wanted to give a bit more information in case this is helpful. Although I am using both youth (waves 1,3,5 and 7) and main survey data (waves 7-13), they do not have to complete every wave. I am interested in certain age groups, so some people may be in that age group and have outcomes in wave 8, some in wave 10, some in wave 12, some in wave 13 etc. I have checked the longitudinal weights from wave 13 that I put in my original message, and unfortunately this will exclude a large portion of my sample. My supervisor suggested possibly using a design weight that is not wave dependent. I wondered if you had any thoughts on this or other suggestions? I can see you mentioned 'pooling', but I am unsure what this is.

Best wishes,

Georgina

*slight amendment to my previous message:

Hi Olena,

Thank you for this. I don't think I explained my research design very well so I wanted to give a bit more information in case this is helpful. Although I am using both youth (waves 1,3,5 and 7) and main survey data (waves 7-13), they do not have to complete every wave as my design is interested in certain age groups in the youth and main survey. So some people may be in the age group of interest (11-13) in wave 1 of the youth survey, or wave 3 etc. and then in the age group of interest (18+) with outcomes in the main survey in wave 8, or in wave 10, or in wave 13 etc. I have checked the longitudinal weights from wave 13 that I put in my original message, and unfortunately this will exclude a large portion of my sample. My supervisor suggested possibly using a design weight that is not wave dependent. I wondered if you had any thoughts on this or other suggestions? I can see you mentioned 'pooling', but I am unsure what this is.

Best wishes,

Georgina

Georgina,

Don't use design weights - these will ignore nonresponse completely.
Yes, you can pool. Read about it here, questions 14 onwards:
https://www.understandingsociety.ac.uk/wp-content/uploads/working-papers/2024-01.pdf

Hope this helps,
Olena