Question

Status:Closed    Asked:Jul 28, 2017 - 10:43 AM

Weighted counts for ATUS-X

Hi,


I have been trying to obtain weighted counts for gender and race using ATUS-X data (I use a pooled file 2003-2016, but in the output below I only show 2010 data so that I could compare the weighted counts to actual counts in population in 2010). I used the "svyset" command and then using "svy: tabulate gender, count" to get the weighted counts. But it doesn't seem to give weighted counts in the total population (the output shows that population size is "88,298,420,349", which seems way too higher than actual population size 309 million in 2010). Could you guide me how to obtain the weighted counts to the population level? Thanks!

Please see the command and output below:


svyset [pw=wt06], sdrweight(rwt06_1 - rwt06_160) vce(sdr)
svy: tab sex, count



. svy: tab sex, count
(running tabulate on estimation sample)

Number of obs = 13,260
Population size = 88,298,420,349
Replications = 160

----------------------
Sex | count
----------+-----------
Male | 4.3e+10
Female | 4.6e+10
|
Total | 8.8e+10
----------------------
Key: count = weighted count


 
Do you have the same question? Follow this Question
 

Staff Answer

avatar

Jeff Bloem

Staff

The sample weights in ATUS data are a bit different than sample weights in other Bureau of Labor Statistics or Census Bureau data. In ATUS, sample weights indicate the number of person-days the respondent represents. In 2003 and 2004, summing the weights of all respondents for a given month yields the number of person-days in that month - which is the total population times the number of days in the month. In 2005 and later, summing the weights of all respondents for a given quarter yields the number of person-days in that quarter - total population times the number of days in the quarter. (More information about all of this can be found in the ATUS User's Guide.) So, this is why you are finding a number that is so much larger than the total population. If you divide the sum of the weights of all respondents by either the number of days in a month or quarter (depending on the sample year), you should calculate a valid estimate of the total population size.

I hope this helps.

 

Aug 02, 2017 - 11:14 AM

0
0
Report it

OTHER QUESTIONS NEEDING ANSWERS

weights for partial sample in March CPS 2014
Change in sample size in 94 and 95 when linking individuals with non-missing ...
When using replicators the database; my tabulated results using the svy comma...
Can I merge basic monthly data with March supplement?
ACS 2010 Hispanic and NonHispanic population data is inconsistent with other ...
Login   |   Register

Recently Active Members

View More »

Share |