NIU (Not in Universe) Cases and Svyset

[Yawo Kokuvi]

Hello all,

I am working on an NHIS dataset that has a lot of missing cases, leading partly to a problem I am trying to resolve (details here: http://www.statalist.org/forums/foru...lation-members).

I am posting a separate question because of its generality.

The dataset has a lot of respondents who were deemed ineligible for certain questions. For example, I am interested in HIV testing, a question that was asked only of sample adults aged 18+.
years. Respondents under 18 yrs are tagged with an NIU code for the HIV and related variables.

Before I do my svyset, do i delete these NIU responses from the data, since my key dependent variable is restricted to only those aged 18+ years or is it better to use the svy subpop option to restrict the sample to only those who are eligible respondents?

Thanks - Yy

[Marcos Almeida]

Hello Yawo,

In order to try to provide an answer for your question, I decided to perform a couple of estimations so as to help you with the reflection about the results:

.   use http://www.stata-press.com/data/r14/nhanes2.dta

. svyset psu [pweight=finalwgt], strata(strata)

      pweight: finalwgt
          VCE: linearized
  Single unit: missing
     Strata 1: strata
         SU 1: psu
        FPC 1: <zero>

. gen highbp2 = highbp

. replace highbp2 = . if race ==1
(9,065 real changes made, 9,065 to missing)

. svy, subpop(if race ==2): logit highbp2 sex age
(running logit on estimation sample)

Survey: Logistic regression

Number of strata   =        30                Number of obs     =       10,013
Number of PSUs     =        60                Population size   =  113,415,086
                                              Subpop. no. obs   =        1,086
                                              Subpop. size      =   11,189,236
                                              Design df         =           30
                                              F(   2,     29)   =        83.52
                                              Prob > F          =       0.0000

------------------------------------------------------------------------------
             |             Linearized
     highbp2 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         sex |  -.1886847   .1284104    -1.47   0.152    -.4509337    .0735642
         age |   .0584824   .0044914    13.02   0.000     .0493098    .0676551
       _cons |  -2.347611    .312217    -7.52   0.000    -2.985243   -1.709979
------------------------------------------------------------------------------
Note: 1 stratum omitted because it contains no subpopulation members.

. drop if race ==1
(9,065 observations deleted)

. svy, subpop(if race ==2): logit highbp2 sex age
(running logit on estimation sample)

Survey: Logistic regression

Number of strata   =        30                 Number of obs     =       1,281
Number of PSUs     =        58                 Population size   =  14,100,211
                                               Subpop. no. obs   =       1,086
                                               Subpop. size      =  11,189,236
                                               Design df         =          28
                                               F(   0,     28)   =           .
                                               Prob > F          =           .

------------------------------------------------------------------------------
             |             Linearized
     highbp2 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         sex |  -.1886847          .        .       .            .           .
         age |   .0584824          .        .       .            .           .
       _cons |  -2.347611          .        .       .            .           .
------------------------------------------------------------------------------
Note: 1 stratum omitted because it contains no subpopulation members.
Note: Missing standard errors because of stratum with single sampling unit.

Hopefully that helps!

Best,

Marcos

[Yawo Kokuvi]

thanks - Marcos. Your illustration makes sense .... so, it is better to restrict the data using the sub-pop command.

Now, just as a final shot, do you have any guidance on merging PSU's with single observations to adjoining ones ?

thanks - Yy

[Marcos Almeida]

Hello Yawo,

I suggest you take a look at the options in svyset for the option - sorry for the pun - singleunit(). The default, as you realized, is singleunit(missing).

Best,

Marcos