Problem with nlcom after svy:mean

[Stuart Morrison]

Hi all,

I am trying to compute a variance equality test with weighted data. I found this very helpful on Stack Exchange:

https://stats.stackexchange.com/ques...ed-survey-data

In which Stas Kolenikov suggests the following code :

webuse nhanes2, clear
gen bpsystol_sq = bpsystol* bpsystol
svy : mean bpsystol*, over( female )
* estimated variance in group 0
nlcom ( _b[bpsystol_sq:0] - _b[bpsystol:0]*_b[bpsystol:0])
* estimated variance in group 1
nlcom ( _b[bpsystol_sq:1] - _b[bpsystol:1]*_b[bpsystol:1])
* equality test
testnl ( _b[bpsystol_sq:1] - _b[bpsystol:1]*_b[bpsystol:1]) ///  
   = ( _b[bpsystol_sq:0] - _b[bpsystol:0]*_b[bpsystol:0])

However, when I run this (and hundreds of variations thereof) I get an error saying equation bpsystol_sq not found.

The coefficients are indeed stored..

 matrix list e(b)

e(b)[1,4]
c.bpsystol@ c.bpsystol@ c.bpsystol~q@ c.bpsystol~q@
0.female 1.female 0.female 1.female
y1 129.92525 124.20266 17255.068 15945.381

But, attempting to reference them by location gives the following:

testnl (e(b)[1,4] - e(b)[1,2]*e(b)[1,2]) = ( e(b)[1,3] - e(b)[1,1]*e(b)[1,1])
invalid '2'

As does various other variations on that theme. I'd be very grateful for any advice.

Cheers,

Stu.

[Rich Goldstein]

it appears that syntax has changed; immediately after estimating your "svy: mean ..." model, try the following to learn how Stata is now referring to the equations:

. svy: mean, coeflegend

Survey: Mean estimation

Number of strata = 31            Number of obs   =      10,351
Number of PSUs   = 62            Population size = 117,157,513
                                 Design df       =          31

------------------------------------------------------------------------------
             |       Mean   Legend
-------------+----------------------------------------------------------------
  c.bpsystol@|
      female |
       Male  |   129.9253  _b[[email protected]]
     Female  |   124.2027  _b[[email protected]]
             |
          c. |
 bpsystol_sq@|
      female |
       Male  |   17255.07  _b[[email protected]]
     Female  |   15945.38  _b[[email protected]]
------------------------------------------------------------------------------

you can now use the above info in setting up your -nlcom- statements; here is the first:

nlcom ( _b[[email protected]] - _b[[email protected]]*_b[c.bpsyst
> [email protected]])

[Stuart Morrison]

Thanks very much indeed. Much appreciated. it's unusual no though for Stata's syntax to change? I confess it didn't occur to me that it might have done as normally one of its selling points is that the code is fully backwards compatible. Perhaps I should start using version!