Unable to replicate Stata 16 code local x = r(table)[1,1] in Stata 15

[Enrique Hernandez]

Dear all,

I have run into a problem trying to replicate some code from Stata 16 in Stata 15. Do you have any idea of what might be going on? I read the update notes and could not identify why this is happening.

Here you have the code that works in Stata 16.

. * Using Stata SE 16.1
.
. sysuse auto.dta, clear
(1978 Automobile Data)

.
. recode rep78 (2=1) (3/5=0)
(rep78: 67 changes made)

.
. probit rep78 mpg headroom

Iteration 0:   log likelihood = -28.552789  
Iteration 1:   log likelihood = -27.844339  
Iteration 2:   log likelihood = -27.837343  
Iteration 3:   log likelihood = -27.837342  

Probit regression                               Number of obs     =         69
                                                LR chi2(2)        =       1.43
                                                Prob > chi2       =     0.4890
Log likelihood = -27.837342                     Pseudo R2         =     0.0251

------------------------------------------------------------------------------
       rep78 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         mpg |  -.0482431    .043335    -1.11   0.266    -.1331782     .036692
    headroom |  -.0784669   .2285802    -0.34   0.731    -.5264758     .369542
       _cons |    .168579   1.340746     0.13   0.900    -2.459235    2.796393
------------------------------------------------------------------------------

.
. local c2_apath_coef = round(r(table)[1,1], .01)

.
. di `c2_apath_coef'
-.05

.
end of do-file

Here the same code that generates an error in Stata 15

. * Using Stata SE 15.1
.  
. sysuse auto.dta, clear
(1978 Automobile Data)

. 
. recode rep78 (2=1) (3/5=0)
(rep78: 67 changes made)

. 
. probit rep78 mpg headroom

Iteration 0:   log likelihood = -28.552789  
Iteration 1:   log likelihood = -27.844339  
Iteration 2:   log likelihood = -27.837343  
Iteration 3:   log likelihood = -27.837342  

Probit regression                               Number of obs     =         69
                                                LR chi2(2)        =       1.43
                                                Prob > chi2       =     0.4890
Log likelihood = -27.837342                     Pseudo R2         =     0.0251

------------------------------------------------------------------------------
       rep78 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         mpg |  -.0482431    .043335    -1.11   0.266    -.1331782     .036692
    headroom |  -.0784669   .2285802    -0.34   0.731    -.5264758     .369542
       _cons |    .168579   1.340746     0.13   0.900    -2.459235    2.796393
------------------------------------------------------------------------------

.  
. local c2_apath_coef = round(r(table)[1,1], .01)
invalid syntax
r(198);

end of do-file

[Joseph Coveney]

Did you see the mistake yet?

If not, then compare the two lines here

local c2_apath_coef = round(r(table)[1,1], .01)
local c2_apath_coef = r(table)[1,1], .01)

[Enrique Hernandez]

Dear Joseph,

Thanks for your reply and for noticing the mistake and divergence between the two codes. I have corrected the mistake in the original post, but the problem persists.

Do you have any idea of what might be going on?

[Stephen Jenkins]

My recollection is that the ability to refer to explicitly-subscripted stored results, as in

 
 r(table)[1,1]

was a feature introduced in Stata 16.

In both Stata 15 and 16 you could (warning -- untested):

matrix T = r(table)
local c2_apath_coef = round(T[1,1], .01)

[Nick Cox]

I don't have access to Stata 15 at present, but in general

local foo = round(whatever, 0.1)

is almost(*) never better than

local foo : di %2.1f whatever

as the first sometimes bites you and the second never should.

Rounding to multiples of 0.1, 0.01, 0.001, ... is a formatting question, not a matter of decimal arithmetic, as most multiples of 0.1, 0.01, 0.001, ... cannot be held in exact binary equivalents.

(*) I can't think of an exception right now, hence the mild fence-sitting.

(This doesn't solve the question posed.)

[Enrique Hernandez]

Thanks Stephen, your solution works perfectly.

I am happy that new feature was introduced in Stata 16.

Thanks Nick for your useful advice on rounding.

[daniel klein]

Also, see function el() as in

local c2_apath_coef = round(el(r(table), 1, 1), .01)

Best
Daniel

[Nick Cox]

Stephen Jenkins is right.

8. Matrix subscripting and extraction are now allowed with r() and e() matrices, including e(b), e(V), and e(Cns). For example,

. sysuse auto
. regress mpg turn trunk
. display e(v)["turn","trunk"]

This information can be found at https://www.stata.com/help.cgi?whatsnew15to16 -- regardless of whether Stata 16 is installed.