Support #667
closed
how best to weight USOC youth data when linking waves for longitudinal analysis
100%
Description
This is a further question for Peter Lynn, following on from his response to my previous question.
I am doing an longitudinal analysis of the young people in USOC who were aged 10-11 at Wave 2, following them through to Wave 4 and Wave 6 and looking at patterns of change in specific questions. As there are no longitudinal weights currently available for the youth questionnaires, what would be the best way to weight this linked data? Would it be best to use the Wave 6 cross-sectional weight? What would be the implications of this for generalising our findings? Thanks for your help.
Updated by Victoria Nolan about 8 years ago
- Status changed from New to In Progress
- Assignee changed from Olena Kaminska to Jane Lakey
- % Done changed from 0 to 10
- Private changed from Yes to No
Dear Jane,
I've passed on your query to Peter and he will reply shortly.
Best wishes, Victoria
Updated by Peter Lynn about 8 years ago
Jane,
The best way would be to start with the wave 2 cross-sectional weight and then make an adjustment to it, to account for attrition at waves 4 and 6. Basically, you would create a 0/1 indicator for each 10/11 year-old at wave 2, to indicate whether they also responded at waves 4 and 6. Then model this indicator on some relevant wave 2 covariates. Use the model predicted values as adjustment factors to the weight. i.e., something like this (where w46out is the 0/1 indicator and x1, x2, etc are your predictor variables):
logistic w46out i.x1 i.x2 ......
pred prob
weight=b_ythscus_xw/prob
Hope that helps,
Peter
Updated by Victoria Nolan about 8 years ago
- Status changed from In Progress to Resolved
- % Done changed from 10 to 100
Updated by Victoria Nolan about 8 years ago
- Status changed from Resolved to Closed