Understanding Society User Support

Apologies, the code I used in (1) is actually:

xtreg fimnlabnet_dv scghq2_dv $controls if wave==1|2 [pw=b_indinus_lw], fe cluster(pidp)
cap drop nonattrition
gen nonattrition = e(sample)
replace nonattrition = . if wave!=2

Freya,

Thank you for your question. I am not sure you need a tailored weight for your analysis. It sounds like you are planning to use a pooled analysis. Can you explain your data set up better? Are you pooling cross-sectional information from 13 waves? Or are you studying any longitudinal aspect, and if so, how many waves at a time are you looking at?

From what I can see in your description your tailored weight does not work, because your resp2 is wrong (all are respondents). Once you have a clear definition of your data structure, you will be able to define your resp2, and your nonresponse model will run.

Hope this helps,
Olena

Hi Olena,

Thank you for your swift response.

I have pooled data from all 13 waves using the 'append' command. So I think you are correct, it's pooled cross-sectional analysis as I'm using xtreg and xtlogit commands to look at how an individuals' mental health impacts their labour market outcomes but I'm not doing this for separate waves or over time. Apologies.

In this case, building off of other queries, I am led to believe the correct weighting approach would be to create a new weight that combines the `w'_indinus_lw weight from each wave, and then re-scale it to account for differences in sample sizes, is this correct?

I have attempted to do this using the following code:

gen weightscaled=0

gen ind=1
sum ind [aw=b_indinus_lw] if wave==2
gen bwtdtot=r(sum_w)

sum ind [aw=c_indinus_lw] if wave==3
gen cwtdtot=r(sum_w)

sum ind [aw=d_indinus_lw] if wave==4
gen dwtdtot=r(sum_w)

etc

replace weightscaled=c_indinus_lw*(bwtdtot/cwtdtot) if wave==3
replace weightscaled=d_indinus_lw*(bwtdtot/dwtdtot) if wave==4
replace weightscaled=e_indinus_lw*(bwtdtot/ewtdtot) if wave==5
replace weightscaled=f_indinus_lw*(bwtdtot/fwtdtot) if wave==6
replace weightscaled=g_indinus_lw*(bwtdtot/gwtdtot) if wave==7
replace weightscaled=h_indinus_lw*(bwtdtot/hwtdtot) if wave==8
replace weightscaled=i_indinus_lw*(bwtdtot/iwtdtot) if wave==9
replace weightscaled=j_indinus_lw*(bwtdtot/jwtdtot) if wave==10
replace weightscaled=k_indinus_lw*(bwtdtot/kwtdtot) if wave==11
replace weightscaled=l_indinus_lw*(bwtdtot/lwtdtot) if wave==12
replace weightscaled=m_indinus_lw*(bwtdtot/mwtdtot) if wave==13

Would you be able to advise if this code is correct please?

One thing I noticed when using the commands:

sum ind [aw=weightscaled] if wave==2
sum ind [aw=weightscaled] if wave==3
sum ind [aw=weightscaled] if wave==4
etc

to check my weight variable, is that sum ind [aw=weightscaled] if wave==2 returns 0 observations and the weight takes a value of 0. I understand why this has occurred however, I am wondering if it would be advisable to add the code replace weightscaled=b_indinus_lw*(bwtdtot/bwtdtot) if wave==2? I have not added it thus far as I was attempting to follow guidance on creating rescaled weights found in previous queries.

Additionally, my code excludes wave 1 as there is no a_indinus_lw. Is there an alternative weight I could use for wave 1 instead?

Moreover, when I attempted to use the weight with the command:

xtreg logincome log_ghq_caseness [aweight=weightscaled], fe vce(cluster pidp)

I got this response from stata:

weight must be constant within pidp
r(199);

Do you have any advice on how to overcome this issue?

Additionally, would it be advisable to do a similar process (i.e., new variable creation and scaling) for psu and strata variables?

Thank you once again for taking the time to respond to my queries, I really appreciate your help.

Kind regards,
Freya

Freya,

Yes, this sounds correct. You would need _xw weights, not _lw weights if you are just pooling cross-sectional information. I will copy my general advice for rescaling below.

Strata and psu variables are constant, and you don't need to change them.

I will forward your message to another collegue to help you with xtreg.

Hope this helps,
Olena

Scaling weights
In pooled analysis and sometimes in other types of analysis you may need to apply an additional scaling to our weights. Our weights have a mean of 1 in each wave, which means that if combined in a pooled analysis the waves with smaller sample size will have a smaller contribution in your analysis. This includes BHPS waves and later waves (as sample size decreases with attrition). Ideally, when combining events / states over 30 years (for example) you want each year to have the same importance. To ensure this follow this example to calculate an additional scaling for your weights.
For example, you are looking at job quality and therefore are pooling information from wave 2, 4, 6 & 8 as these are the waves when the questions are asked. Here is how to create a scaled weight for this analysis.

ge weightscaled=0
replace weightscaled=b_indpxub_xw if wave=2

ge ind=1
sum ind [aw=b_indpxub_xw] if wave=2
gen bwtdtot=r(sum_w)
sum ind [aw=d_indpxub_xw] if if wave=4
gen dwtdtot=r(sum_w)
sum ind [aw=f_indpxub_xw] if if wave=6
gen fwtdtot=r(sum_w)
sum ind [aw=h_indpxub_xw] if if wave=8
gen hwtdtot=r(sum_w)

replace weightscaled=d_indpxub_xw*(bwtdtot/dwtdtot) if wave=4
replace weightscaled=f_indpxub_xw*(bwtdtot/fwtdtot) if wave=6
replace weightscaled=h_indpxub_xw*(bwtdtot/hwtdtot) if wave=8

You can double check by looking at the sum of ind with weightscaled for each wave – it should be the same.
sum ind [aw=weightscaled] if wave==2
sum ind [aw=weightscaled] if wave==4
sum ind [aw=weightscaled] if wave==6
sum ind [aw=weightscaled] if wave==8