# Re: st: RE: Lincom after sepov

 From "Joao Ricardo F. Lima" To statalist@hsphsun2.harvard.edu Subject Re: st: RE: Lincom after sepov Date Tue, 24 Jun 2008 20:40:14 -0300

```Dear Nick,

sorry this question, my poor knowledge of Stata or the english language, oks?

But whatīs svy sawy?

I know what is svy, of course.

Best,

Joao Ricardo

2008/6/24, Nick Cox <n.j.cox@durham.ac.uk>:
> Thanks for the correction to your previous post.
>
> The same point applies. It is not -sepov- doing this, but whatever it calls.
>
> Someone else -svy-savvy may be able to help with your main problem.
>
> Nick
> n.j.cox@durham.ac.uk
>
> Joao Ricardo F. Lima
>
> Thanks Nick,
>
> actually, sepov store the values of FGT in e(b), not in e(V). See:
>
> sepov renagr_nag rennag if famtipo == 1 [w=pesopes2], p(120)
> strata(strat) psu(psu) ci meff
> (sampling weights assumed)
>
> Poverty measures for the variable renagr_nag: (unlabeled)
>
> Survey mean estimation
>
> pweight:  pesopes2                                Number of obs    =
> 849
> Strata:   strat                                   Number of strata =
> 43
> PSU:      psu                                     Number of PSUs   =
> 153
>                                                   Population size  =
> 394883.36
>
> ------------------------------------------------------------------------------
>     Mean |   Estimate    Std. Err.   [95% Conf. Interval]        Meff
> ---------+--------------------------------------------------------------------
>       p0 |   .1953685    .0173466    .1609916    .2297453    1.644852
>       p1 |   .0671076    .0064495    .0543262     .079889    1.270128
>       p2 |   .0327589    .0037762    .0252754    .0402425    1.198951
> ------------------------------------------------------------------------------
>
>
> Poverty measures for the variable rennag: (unlabeled)
>
> Survey mean estimation
>
> pweight:  pesopes2                                Number of obs    =
> 849
> Strata:   strat                                   Number of strata =
> 43
> PSU:      psu                                     Number of PSUs   =
> 153
>                                                   Population size  =
> 394883.36
>
> ------------------------------------------------------------------------------
>     Mean |   Estimate    Std. Err.   [95% Conf. Interval]        Meff
> ---------+--------------------------------------------------------------------
>       p0 |   .0649527    .0079664    .0491652    .0807402    .9895778
>       p1 |   .0218262    .0033303    .0152263    .0284261    1.086596
>       p2 |   .0098536     .002091    .0057097    .0139976    1.136367
> ------------------------------------------------------------------------------
>
> . matrix list e(b)
>
> e(b)[1,3]
>            p0         p1         p2
> y1   .0649527  .02182618  .00985364
>
>
> I want to compare if 0.0649527 = 0.1953685 statistically, but donīt
> known how to do this.
>
> 2008/6/24 Nick Cox <n.j.cox@durham.ac.uk>:
>> Standard advice for this list (more usually ignored than not) is to say
>> where user-written stuff you are using comes from.
>>
>> -sepov- has this source
>>
>> STB-51  sg117 . . Robust std errors for Foster-Greer-Thorbecke poverty
>> indices
>>        (help sepov if installed) . . . . . . . .  D. Jolliffe and A.
>> Semykina
>>        9/99    pp.34--36; STB Reprints Vol 9, pp.200--203
>>        provides estimates of standard errors for the Foster-Greer-
>>        Thorbecke class of poverty indices; handles complex survey
>>        designs including stratification and multi-stages
>>
>> and evidently the authors have moved since then.
>>
>> A Google for Dean Jolliffe seems to yield a direct hit. I don't know Dean
>> and have no way of knowing whether he still supports -sepov-.
>>
>> But the code remains accessible. -viewsource sepov.ado- and you can look
>> at the code and have a stab at answering your own question. You will see
>> that your statement is not quite correct. -epov- is not eclass and does
>> not store anything in e(). Whatever e(V) you see is left over from
>> -svymean-, or so I guess. This may help you move forward.
>>
>> Nick
>> n.j.cox@durham.ac.uk
>>
>> Joao Ricardo F. Lima
>>
>> I'm writing because I want an advice. I'm using sepov to get FGT
>> indexes for two variables. After estimation I would like to compare if
>> the results are statistical different. I am trying to use "lincom",
>> but unsucessfull because sepov stores in e(V) only the estimation for
>> the last variable, isnīt it?
>>
>> I am doing this:
>>
>> sepov var1 var2 , p(pl) psu (psu) strat (strat) ci meff
>>
>> And I would like something like this
>>
>> lincom p0[var1] - p0[var2]
>>
>> How can I do this? Is there another way to test if the values of FGT
>> indexes are the same?
>>
>> I wrote to sepov authors, but their email address change.
>>
>> The error that the other server returned was: 550 550 5.1.1 Mailbox
>> <anastassia.semykina@cerge.cuni.cz> does not exist (state 14).
>> The error that the other server returned was: 550 550 5.1.1 Mailbox
>> <dean.jolliffe@cerge.cuni.cz> does not exist (state 14).
>
> *
> *   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/
>

--
-------------------------------
Joao Ricardo Lima
Professor
UFPB-CCA-DCFS
+553138923914
-------------------------------

*
*   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/
```