Support #1678
Weights for the linked COVID-19 Youth data
100%
Description
Hello!
Hope this email finds you well.
I am currently working with COVID-19 youth data and I have a question about weights. I have already read the user guide you previously sent me, however, I still have a question.
1) I have merged COVID youth wave 4 (July 2020) and wave 8 (March 2021). There are two weights available in this datasets (cd_betayth_xw and ch_betayth_xw). Do I need to use them both? Or just one of them? And do I need to use any other weights when I am merging to COVID waves for youth?
2) Then I have another dataset - I have merged (1) COVID youth wave 4 (July 2020), (2) COVID youth wave 8 (March 2021) and (3) baseline UKLHS Wave 10 (2018-19). Which weight I need to use in this case?
Thank you and looking forward to hearing from you!
Kind regards,
Irina
Files
Updated by Understanding Society User Support Team about 2 years ago
- Status changed from New to In Progress
Many thanks for your enquiry. The Understanding Society team is looking into it and we will get back to you as soon as we can.
We aim to respond to simple queries within 48 hours and more complex issues within 7 working days.
Best wishes,
Understanding Society User Support Team
Updated by Understanding Society User Support Team about 2 years ago
Dear Irina,
Is your analysis cross-sectional or longitudinal?
Best wishes,
Understanding Society User Support Team
Updated by Understanding Society User Support Team about 2 years ago
- Status changed from In Progress to Feedback
- % Done changed from 0 to 10
- Private changed from Yes to No
Updated by Understanding Society User Support Team about 2 years ago
- Category changed from Data linkage and consents to COVID-19
Updated by Irina Kolegova about 2 years ago
As I am using the same respondents who took part in the pre-covid wave, the 4th Covid wave and the 8th Covid wave I think that this analysis is longitudinal. But please correct me if I am wrong.
Best wishes,
Irina
Updated by Irina Kolegova about 2 years ago
Let me show my steps:
1) I merged youth who participated in Wave 4 (July 2020) and Wave 8 (March 2021) and kept those who matched observations (N=821):
use "/Users/irina/Desktop/Thesis/2022/Data/UKDA-8644-stata/stata/stata13_se/cd_youth_p.dta"
merge 1:1 pidp_c using "/Users/irina/Desktop/Thesis/2022/Data/UKDA-8644-stata/stata/stata13_se/ch_youth_p.dta", nogenerate keep(match))
2) then I merged it to the baseline (wave 10 UKHLS) and also kept matched observations (N=531):
merge 1:1 pidp using "/Users/irina/Desktop/Thesis/2022/Data/UKDA-6614-stata/stata/stata13_se/ukhls_w10/j_youth.dta", nogenerate keep(match)
3) then I want to compare how SDQ score changed over time. For this I want to test absolute differences in SDQ scores between wave 10 UKHLS (2018/19) and wave 4(July 2020) samples using linear regression for continuous outcomes
Updated by Understanding Society User Support Team about 2 years ago
Could you please also include the syntax for the regression part?
Best wishes,
Understanding Society User Support Team
Updated by Irina Kolegova about 2 years ago
I have just realised that I do not know which dependent variable I should use for my regression.
I was thinking to get a table like this. Do you know how can I do it?
Updated by Irina Kolegova about 2 years ago
Hello,
I have finally managed to figure out the syntax for the regression part.
Here is my syntax:
// July 2020 + March 2021
use "/Users/irina/Desktop/Thesis/2022/Data/UKDA-8644-stata/stata/stata13_se/cd_youth_p.dta"
merge 1:1 pidp_c using "/Users/irina/Desktop/Thesis/2022/Data/UKDA-8644-stata/stata/stata13_se/ch_youth_p.dta", nogenerate keep(match)
rename pidp_c pidp
save "/Users/irina/Desktop/Thesis/2022/Data/youth data/cd_ch_youth_p.dta", replace
- Adding baseline Wave 10 UKHLS (2019) to July 2020 + March 2021
use "/Users/irina/Desktop/Thesis/2022/Data/youth data/cd_ch_youth_p.dta"
merge 1:1 pidp using "/Users/irina/Desktop/Thesis/2022/Data/UKDA-6614-stata/stata/stata13_se/ukhls/i_youth.dta", nogenerate keep(match)
//// Renaming variables and transforming wide data to long ////
rename cd_* 2020
rename ch_ 2021
rename i_ *2018
//Reshaping data
reshape long ypsdqtd_dv ypsdqes_dv ypsdqcp_dv ypsdqha_dv ypsdqpp_dv ypsdqps_dv, i(pidp) j(year)
label variable ypsdqtd_dv "Total SDQ"
label variable ypsdqes_dv "Emotional Symptoms"
label variable ypsdqcp_dv "Conduct Problems"
label variable ypsdqha_dv "Hyperactivity/Inattention"
label variable ypsdqpp_dv "Peer Relationship Problems"
label variable ypsdqps_dv "Prosocial Behaviour"
// Testing changes in TOTAL mean scores
regress ypsdqtd_dv i.year if year==2020 | year==2018
// Testing changes in subscale mean scores
*Emotional S
regress ypsdqes_dv i.year if year==2020 | year==2018
*Conduct Problems
regress ypsdqcp_dv i.year if year==2020 | year==2018
*Hyperactivity/Inattention
regress ypsdqha_dv i.year if year==2020 | year==2018
*Peer Relationship Problems
regress ypsdqpp_dv i.year if year==2020 | year==2018
**Prosocial Behaviour
regress ypsdqps_dv i.year if year==2020 | year==2018
Updated by Irina Kolegova about 2 years ago
Updated by Understanding Society User Support Team about 2 years ago
- % Done changed from 10 to 80
Hi Irina,
It seems that you want to estimate trend in these SDQ summary scales, rather than within person change in these scores even though you have restricted the data to those who have responded to Wave 10, Covid Waves 4 & 8. But as you are not estimting within person change, you don't need to restrict the sample to those who have responded in all 3 waves and can use cross-sectional weights for the respective waves.
Also note that even though you are using "year" as an explanatory variable, the value of "2018" is actually "2018-20" as the interviews for Wave 10 were mostly in 2018 & 2019 plus a few in 2020. You could restrict the Wave 10 observations to only those who were interviewed in 2018 & 2019 so that these observations represented pre-pandemic.
When producing weighted estimates, you can use svyset to take into account weights as well as the complex survey design.
Best wishes,
Understanding Society User Support Team
Updated by Irina Kolegova about 2 years ago
Dear Understanding Society User Support Team,
Thank you for your response.
Just a quick remark - I am not using UKHLS Wave 10. For a baseline I use UKHLS Wave 9 (2017-2018).
It sounds good that I don't have to restrict the sample to those who have responded in all 3 waves. Could you please show me in my code where I should apply the "cross-sectional weights for the respective waves"? Do I need to apply weights only when I run regressions? Or should I apply weights in the beginning of the analysis once I've merged the 3 waves?
Could you please also explain me how can I "use svyset to take into account weights as well as the complex survey design"?
Thank you and looking forward to hearing from you
Best,
Irina
Updated by Understanding Society User Support Team about 2 years ago
Sorry about that - same principle applies for baseline using Wave 9.
After you have put the data together (as you have shown above) create a variable called weight and replace it with the different wave specific youth cross-sectional weights, depending on the wave.
As weights are about producing unbiased population estimates, apply them when producing population estimates. Here is some infomration about using weights and complex survey design in the main survey user guide: https://www.understandingsociety.ac.uk/documentation/mainstage/user-guides/main-survey-user-guide/how-to-use-weights-analysis-guidance-for-weights-psu-strata
Hope this helps
Updated by Irina Kolegova about 2 years ago
Hello! I have applied the weights for the youth as you recommended, could you please tell me if I have done it correctly? (attached screenshot)
Updated by Irina Kolegova about 2 years ago
and do I need to keep only matched observations (when I am merging waves in the beginning) in this case?
Updated by Understanding Society User Support Team about 2 years ago
- Assignee changed from Irina Kolegova to Alita Nandi
Updated by Understanding Society User Support Team about 2 years ago
- Assignee changed from Alita Nandi to Understanding Society User Support Team
As your objective i to estimate trends, you don't need to restrict to cases that responded in all 3 waves.
Updated by Understanding Society User Support Team about 2 years ago
looked t the screenshop - yes, weights created and use of svyset is ok
Updated by Understanding Society User Support Team about 2 years ago
- Status changed from Feedback to Resolved
- % Done changed from 80 to 100