I would like to generate population-level estimates using the AHTUS. Is it possible to use the available survey weights to do that? I’ve researched the recwght variable and it does not appear to provide the necessary information. Thank you!
The AHTUS user guide linked from RECWGHT provides a bit more information on this topic (see the “Weights” section). INFLTWT is the sample weight that inflates RECWGHT to the national population.
1 Like
Hi Kari,
Thank you so much for the response a couple of years ago, it was very helpful! I still have some questions about calculating population-level estimates of total participants in activities in AHTUS. Using infltwt divided by 365 days I obtain numbers that are relatively plausible but they are below the weighted estimates using ATUS. Why is that the case? See below an example:
Using ATUS:
svyset caseid [pw=wt06]
svy: mean sex if year==2018
Population size estimate = 95,591,791,615, which divided by 365 results in about 262 million people.
Using AHTUS:
svyset ident [pw=infltwt]
svy: mean sex if year==2018
Population size estimate = 63,931,201,658, which divided by 365 results in about 172 million people.
See detailed output below. I’m puzzled by this discrepancy. Shouldn’t the two estimates be the same? Which one is correct?
Thanks!
Pilar
OUTPUT using ATUS
. svy: mean sex if year==2018
(running mean on estimation sample)
Survey: Mean estimation
Number of strata = 1 Number of obs = 9,593
Number of PSUs = 9,593 Population size = 95,591,791,615
Design df = 9,592
| Linearized
| Mean std. err. [95% conf. interval]
-------------±-----------------------------------------------
sex | 1.515828 .0070061 1.502095 1.529562
OUTPUT using AHTUS
. svy: mean sex if sample==2018
(running mean on estimation sample)
Survey: Mean estimation
Number of strata = 1 Number of obs = 9,593
Number of PSUs = 9,593 Population size = 63,931,201,658
Design df = 9,592
| Linearized
| Mean std. err. [95% conf. interval]