Re: st: Why does one command "lose" observations compared to another command?

[Richard Williams <Richard.A.Williams.5@ND.edu>]

At 02:16 PM 1/22/2006, Suzy wrote:
> If the mecwt4 variable loses some observations (which is certainly 
> possible), why didn't it also lose those observations for the svy 
> command?  In other words, shouldn't the svy command output have been =
> Subpop. no. obs = 10,568   rather than the 11,441?
>
> Thank you for helping troubleshoot the problem!
Does mecwt4 include weights of zero?  I am guessing that may be 
why.  svy still counts cases with 0 weights, but mean with the weight 
parameter does not.  Here is a pseudo-replication of your problem.  Rich

. webuse nhanes2f

. * subpop size is 9090 in both svy, non-svy analysis
. svy, subpop(if age>=25): mean age, over(race)
(running mean on estimation sample)

Survey: Mean estimation

Number of strata =      31          Number of obs    =   10337
Number of PSUs   =      62          Population size  = 1.2e+08
                                    Subpop. no. obs  =    9090
                                    Subpop. size     = 1.0e+08
                                    Design df        =      31

        White: race = White
        Black: race = Black
        Other: race = Other

--------------------------------------------------------------
             |             Linearized
        Over |       Mean   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
age          |
       White |   45.91354    .308088      45.28519    46.54189
       Black |   44.20919   .5337285      43.12064    45.29773
       Other |   44.03196   2.201001      39.54299    48.52093
--------------------------------------------------------------

. mean age [pweight = finalwgt] if age>=25, over(race)

Mean estimation                     Number of obs    =    9090

        White: race = White
        Black: race = Black
        Other: race = Other

--------------------------------------------------------------
        Over |       Mean   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
age          |
       White |   45.91354    .170648      45.57903    46.24805
       Black |   44.20919   .4886224      43.25138      45.167
       Other |   44.03196   1.073968      41.92674    46.13718
--------------------------------------------------------------

. *************************************
. replace finalwgt = 0 in 1/10
(10 real changes made)

. * We've changed some weights to 0 - but subpop size is still 9090 in svy
. svy, subpop(if age>=25): mean age, over(race)
(running mean on estimation sample)

Survey: Mean estimation

Number of strata =      31          Number of obs    =   10337
Number of PSUs   =      62          Population size  = 1.2e+08
                                    Subpop. no. obs  =    9090
                                    Subpop. size     = 1.0e+08
                                    Design df        =      31

        White: race = White
        Black: race = Black
        Other: race = Other

--------------------------------------------------------------
             |             Linearized
        Over |       Mean   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
age          |
       White |   45.90365    .304178      45.28327    46.52402
       Black |   44.19801   .5404275       43.0958    45.30022
       Other |   44.03196   2.201001      39.54299    48.52093
--------------------------------------------------------------

. * But Mean command now reports a smaller subpop size of 9081
. mean age [pweight = finalwgt] if age>=25, over(race)

Mean estimation                     Number of obs    =    9081

        White: race = White
        Black: race = Black
        Other: race = Other

--------------------------------------------------------------
        Over |       Mean   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
age          |
       White |   45.90365   .1707149      45.56901    46.23829
       Black |   44.19801   .4886391      43.24017    45.15585
       Other |   44.03196   1.073968      41.92674    46.13718
--------------------------------------------------------------


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/