Weighting youth data and average parental characteristics
Dear support team,
I have another query regarding the use of weights in the youth data.
I am looking at reported happiness among young people at Waves 1 and Waves 5. I generally want to examine happiness in this group but also want to test if there is a change over time.
So my first question is
(1) If I want to control for parental characteristics(for example: SF12-physical and mental illness, parents age) in my mode1s, and I am using average parental characteristics (i.e. from the mother and father). Should I use cross sectional weights for the youth or their parents at wave 1?
(2) And does the same principle apply at wave 5?
(3) Am I correct in assuming that if I want to examine change at wave 5 then I should use a longitudinal weight?
Thank you for your time.
#1 Updated by Victoria Nolan almost 4 years ago
- Category set to Weights
- Status changed from New to In Progress
- Assignee set to K.Samantha Russell Jonsson
- % Done changed from 0 to 10
- Private changed from Yes to No
I have added Peter Lynn as a watcher for this post and he will be able to get back to you about your weighting query.
Best wishes, Victoria.
#2 Updated by Peter Lynn almost 4 years ago
1) If your units of analysis are the children, and you are treating the parental characteristics as attributes of the children, then use the youth weight;
3) Depends what you mean by change. If your dependent variable is the difference between, say, happiness at wave 5 and happiness at wave 1, then this can only be measured in the youth questionnaire for people aged 10 or 11 at wave 1. I think you would have to use the wave 1 cross-sectional youth weight and then make an adjustment for attrition amongst these 10-11 year-olds by wave 5.
#5 Updated by Peter Lynn almost 4 years ago
Create a data set consisting of wave 1 respondents aged 10 or 11 and a 0/1 indicator of whether or not they also responded at wave 5. Model this indicator based on relevant respondent characteristics from wave 1 (youth qre, hhd qre, hhd grid) (e.g. a logistic regression). This will give you a predicted probability for each wave 1 respondent of wave 5 response. Call this P. You then need to adjust the wave 1 cross-sectional weight for all the cases that can be included in your analysis (i.e. completed youth qre at both w1 and w5) by multiplying it by 1/P.