Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: svy: total command


From   Steve Samuels <sjsamuels@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: svy: total command
Date   Wed, 21 Nov 2012 10:40:20 -0600

Rebecca:

To get the total based on the two data sets, ordinarily, I would say: 
1) append the two data sets;
2) use the same -svyset- statement and -total- commands that you used in the original data. 

The  numbers in your results don't correspond to those you reported on Nov 5

"- Regarding the discrepancies in the number of PSUs; I have different
number, for example, in the file that collects variables about
remittances from household members (PSUs=191) and in the file that
collects the same variables about remittances from friends (PSUs=197).
This may suggests that in some PSUs, hhs that receive remittances from
household members were not sampled."


But the -total- statement in the second data set below shows only 90 PSUs, not
197.   If the coverage of the "friend" remittance information is incomplete, then I
suggest that you report the separate totals and the combined total, but with a note
that the total might be wrong.

Steve


On Nov 20, 2012, at 6:18 AM, Rebecca Pietrelli wrote:

Dear Steve,

first of all, I am very sorry.
My question was inaccurate and I know that time is precious!

The survey I am using collects information on households residing in Uganda.

Sampling design:
A two-stage stratified sample design was adopted. First stage:
Selection of  Enumeration Areas (ea), proportionally done on the basis
of the number of households according to Uganda Household Survey. Also
the selection was done separately for urban and rural  areas. Second
stage: systematic sampling procedure was adopted in order to select
(in each ea) 4 households with an international migrant, 3 households
with one or more internal migrant and 3 with no migrants.

In one data file, for each household, the list of household members
living outside the household is reported, with all the information
about each migrant, including remittances behaviour.
In the second data file, for each household, the list of non-household
members who remit to the household is reported, with all the
information about each migrant.

The aim of my analysis it to compute the total amount of remitters
(and remittances), both former household members and non-household
members (namely household members and friends), distinguishing
internal and international remitters.


I used the following commands in both data files.

. svyset earea [pw=hhweight], strata(stratum)

     pweight: hhweight
         VCE: linearized
 Single unit: missing
    Strata 1: stratum
        SU 1: earea
       FPC 1: <zero>

I have generated a dummy variable, labelled internal, = 1 if the
migrant has migrated within Uganda and = 0 if the migrant has migrated
out of Uganda.
Also I have generated a dummy variable, labelled dremit1, = 1  if the
migrant remits and zero otherwhise.

I used the following command to compute the total number of remitters
in each data files (hh members and friends), distinguishing internal
(1) and international(0):

First file: former household members:

. svy, over(internal): total dremit1
(running total on estimation sample)

Survey: Total estimation

Number of strata =       2         Number of obs    =     1645
Number of PSUs   =     191         Population size  =  4942119
                                  Design df        =      189

           0: internal = 0
           1: internal = 1

--------------------------------------------------------------
            |             Linearized
       Over |      Total   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
dremit1      |
          0 |   233898.1   29978.48      174762.7    293033.5
          1 |    1064315   131143.6      805622.1     1323009
--------------------------------------------------------------

Second file: Non-household members:

. svy, over(internal): total dremit1
(running total on estimation sample)

Survey: Total estimation

Number of strata =       2          Number of obs    =     217
Number of PSUs   =      90          Population size  =  564846
                                   Design df        =      88

           0: internal = 0
           1: internal = 1

--------------------------------------------------------------
            |             Linearized
       Over |      Total   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
dremit1      |
          0 |   101800.5   22591.33         56905      146696
          1 |   324690.8   66555.15      192426.4    456955.2
--------------------------------------------------------------

Finally, I think to sum the total numbers of remitters (household
members and non-household members), separating internal and
international migrants. In other words, the results obtained from the
two tables above.
Also I have applied the same procedure to compute the total amount of
remittances.

Is this procedure (working on two data files) correct?

The weights are at household level and I am computing the total number
of remitters. Is there a problem of bias, like here? If yes, How can I
solve it?

http://www.stata.com/statalist/archive/2010-09/msg00445.html

Thank you and sorry again!

Rebecca








2012/11/20 Steve Samuels <sjsamuels@gmail.com>:
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index