| population size [message #12202] |
Sat, 08 April 2017 14:20  |
catherino222
Messages: 4
Registered: February 2017
Location: London ON
|
Member |
|
|
Hello,
I have a question about the population size in regression outputs. I used the code below for applying weights and survey design in Stata.
egen clusters=group(v021), label
egen strata=group(v024 v025), label
svyset cluster [pweight=v005], strata(strata) singleunit(centered)
My data is Cameroon DHS 2011(15,426 women in sample).
My output, however, shows a very large population size. I thought the population size was suppose to reflect the population of women in the country.
Also, will the population size double if 2 datasets (2004 and 2011)for the same country are pooled together?
Here is the output for CDHS 2011:
Number of strata = 22 Number of obs = 15426
Number of PSUs = 578 Population size = 15425999854
Design df = 556
F( 1, 556) = 127.68
Prob > F = 0.0000
------------------------------------------------------------ ------------------
| Linearized
v301 | Odds Ratio Std. Err. t P>|t| [95% Conf. Interval]
-------------+---------------------------------------------- ------------------
v024 | 1.131586 .0123798 11.30 0.000 1.107528 1.156166
_cons | 7.975635 .9747392 16.99 0.000 6.273487 10.13962
------------------------------------------------------------ ------------------
thank you
Catherine
|
|
|
|
| Re: population size [message #12214 is a reply to message #12202] |
Tue, 11 April 2017 10:46  |
Bridgette-DHS
Messages: 3208
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS Stata Specialist, Tom Pullum:
Your svyset is correct. The weight variable, v005, includes a factor of 1000000 (one million) just so you can ignore any decimal places. Apart from that factor, the sum of the weights is the unweighted sample size, 15426. What is given as the "population size" includes that factor of one million, but Stata has done the weighting correctly.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
