Hello Colleagues!
I'm reaching out with what I gather is a fairly basic inquiry. I'm working with cross-sectional survey data (same in which overall response rates are substantially variable. These survey data have now been collected in three waves (2015, 2018 and 2021, for what it is worth). One sub-population- identified by variable X- has demonstrated importantly distinct attitudes and experiences in all surveys as compared with the entire population- on average. Importantly, the proportion of the overall sample made up by this sub-population has increased significantly over the three waves.
It seems clear to me that the answer to this problem is to create a weight for this subpopulation and apply it to all such analyses, but it's not clear to me:
(1) whether the weight should be the baseline proportion of the sample in wave 1 or the actual proportion of the population.
(2) how best to approach creating and applying these weights given my lack of experience with their creation.
I'm happy to provide more information if it seems helpful, but this does strike me as a pretty simple issue for which many have a great deal of experience and thus might be easy to quickly apprehend and respond.
Deep gratitude to folks for taking the time to respond to this inquiry.
I'm reaching out with what I gather is a fairly basic inquiry. I'm working with cross-sectional survey data (same in which overall response rates are substantially variable. These survey data have now been collected in three waves (2015, 2018 and 2021, for what it is worth). One sub-population- identified by variable X- has demonstrated importantly distinct attitudes and experiences in all surveys as compared with the entire population- on average. Importantly, the proportion of the overall sample made up by this sub-population has increased significantly over the three waves.
- In the first wave/year, the sample size of X subpopulation was slightly smaller than its proportion of the overall population.
- In the second wave/year, the sample size of X subpopulation was slightly larger than its proportion of the overall population.
- In the third wave/year, sample size of X subpopulation was very substantially larger than its proportion of the overall population.
It seems clear to me that the answer to this problem is to create a weight for this subpopulation and apply it to all such analyses, but it's not clear to me:
(1) whether the weight should be the baseline proportion of the sample in wave 1 or the actual proportion of the population.
(2) how best to approach creating and applying these weights given my lack of experience with their creation.
I'm happy to provide more information if it seems helpful, but this does strike me as a pretty simple issue for which many have a great deal of experience and thus might be easy to quickly apprehend and respond.
Deep gratitude to folks for taking the time to respond to this inquiry.
