Support #1838
open
Weights when using Data from waves with gaps
Added by Johanna Pauliks almost 2 years ago.
Updated about 1 year ago.
Description
Greetings,
I'm currently unsure which weights I should use for my analysis. I'm interested in the effect of flexible work arrangements on wages and questions regarding FWAs were only asked in wave 2, 4, 6, 8 and 10. Furthermore, I want to use information from all of these waves and run a random effects panel model in stata. Meaning, I am not interested in changes over time, but I cluster individuals based on time. I first thought, that I could simply use the longitudinal weight from wave 10 and copy it to every wave. But after studying the forum and guides on weighting, I found out that this would exclude everyone who dropped out before wave 10, am I right? Could I simply use cross-sectional weights of each wave that I use, or do I need to create my own weight? What would you advise in this situation?
Best regards
Johanna
Johanna,
Could you explain a bit more the set up of your dataset, please? Do you have multiple observations per person, for example?
Thank you,
Olena
Olena Kaminska wrote in #note-1:
Johanna,
Could you explain a bit more the set up of your dataset, please? Do you have multiple observations per person, for example?
Thank you,
Olena
Dear Olena,
thank you for your quick reply. Yes, I use multiple observations per person. More specifically, I want to use all observations available per Person from wave 2, 4, 6, 8 and 10 (as this is where my independent variable of interest was collected) and run a random effects regression model to address clustering of the observation within Participants over time. Even though my research question is cross-sectional in nature, I would still like to use all data available, as some people do not work from home during the entirety of the panel, but only at one or two waves, for example. My main variables of interest are from the adult main interview (not proxy), though I want to use some control variables on the Household level.
I hope my explanation was sufficient.
Best regards
Johanna
Johanna,
Sounds like you are studying not people but events / states (that can differ over time). Sounds like you don't need a longitudinal weight for your analyses, and should go with cross-sectional weights relevant to each wave.
Hope this helps,
Olena
- Status changed from New to Feedback
- % Done changed from 0 to 50
- Private changed from Yes to No
- Status changed from Feedback to Resolved
- % Done changed from 50 to 100
Also available in: Atom
PDF
Updated by 
Updated by 
Updated by