Syntax for lincom won't work in Stata 16 but works fine in Stata 15

[Lauren Copeland]

Hello all. I'm a professor at Baldwin Wallace. On Thursday, I tried to run a lab with my class from Pollock's Stata Companion (Chapter 6) to compare means. This lab runs fine in Stata v. 15, but it will not run in Stata v. 16.

This is the command
lincom [egal_scale]Male - [egal_scale]Female

It runs perfectly fine in Stata 15, which I have on my home computer (see below).

When I run it in Stata 16, I get an error message that says the variable was not recognized as an equation. All my students ran into this problem, too.

Needless to say, it was a little embarrassing, and 'googling' the error isn't doing the trick.

Any suggestions?

Thank you!
-Lauren



. mean egal_scale [pw=nesw], over(female)

Mean estimation Number of obs = 5,456

Male: female = Male
Female: female = Female

--------------------------------------------------------------
Over | Mean Std. Err. [95% Conf. Interval]
-------------+------------------------------------------------
egal_scale |
Male | 13.69738 .1288302 13.44482 13.94993
Female | 14.44775 .1250026 14.2027 14.69281
--------------------------------------------------------------

.
. lincom [egal_scale]Male - [egal_scale]Female

( 1) [egal_scale]Male - [egal_scale]Female = 0

------------------------------------------------------------------------------
Mean | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | -.7503768 .1795073 -4.18 0.000 -1.102283 -.3984709
------------------------------------------------------------------------------

[Carlo Lazzaro]

Lauren:
welcome to this forum.
I can replicate your problem with one of the traditional Stata database:

. use "C:\Program Files (x86)\Stata15\ado\base\a\auto.dta"
(1978 Automobile Data)

. mean price, over(foreign)

Mean estimation                   Number of obs   =         74

     Domestic: foreign = Domestic
      Foreign: foreign = Foreign

--------------------------------------------------------------
        Over |       Mean   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
price        |
    Domestic |   6072.423   429.4911      5216.449    6928.398
     Foreign |   6384.682   558.9942      5270.608    7498.756
--------------------------------------------------------------

. mat list e(b)

e(b)[1,3]
           0b.         1.          
      foreign    foreign      _cons
y1          0  312.25874  6072.4231

. lincom [ price ]Domestic - [ price ]Foreign

 ( 1)  [price]Domestic - [price]Foreign = 0

------------------------------------------------------------------------------
        Mean |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         (1) |  -312.2587   704.9376    -0.44   0.659    -1717.197     1092.68
------------------------------------------------------------------------------

.




. use "C:\Program Files\Stata16\ado\base\a\auto.dta"
(1978 Automobile Data)

.  mean price, over(foreign)

Mean estimation                      Number of obs   =         74

-----------------------------------------------------------------
                |       Mean   Std. Err.     [95% Conf. Interval]
----------------+------------------------------------------------
c.price@foreign |
      Domestic  |   6072.423   429.4911      5216.449    6928.398
       Foreign  |   6384.682   558.9942      5270.608    7498.756
-----------------------------------------------------------------

. mat list e(b)

e(b)[1,2]
      c.price@   c.price@
    0.foreign  1.foreign
y1  6072.4231  6384.6818

. lincom [ price ][email protected] [ price ][email protected]
equation [price] not found
r(303);

.

However, immediately after my reply, I've noticed that I should have been more careful in reading -mean- entry examples in Stata .pdf manual, as everything boils down to a variation in coding between 15 and 16 releases:

. lincom [email protected] [email protected]

 ( 1)  [email protected] - [email protected] = 0

------------------------------------------------------------------------------
        Mean |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         (1) |  -312.2587   704.9376    -0.44   0.659    -1717.197     1092.68
------------------------------------------------------------------------------

.

The gist of the story is that William Lisowski is >100% right when he encourages a careful (and repeated) reading of the Stata material.

[Lauren Copeland]

Thank you! I also reached out to Stata Technical Support, and this was their response:

From: Stata Technical Support <[email protected]>
Sent: Wednesday, March 4, 2020 11:06 AM
To: Lauren Copeland
Subject: EXT: Re: Lincom in Stata 16

________________________________

Dear Lauren,

Thank you for contacting us.

This is an improvement we made in Stata 16 for -mean- and a couple of
other commands. In -help whatsnew15to16-, you will find the following
information

4. Estimation commands mean, proportion, ratio, and total now
allow factor-variable notation. Old behavior is preserved under
version control. We mention that despite factor variables being
a superset of varlists because variables in over() are now
interacted, which they always were, but the use of factor
variables changes (improves) the names under which results are
stored. Old result names are preserved under version control.

Now factor variables allow more complex combinations in -mean- and more
post-estimations such as -marginsplot- are available after -mean-.

We can see the labels for the mean estimates are changed. That's why
-lincom- cannot find the stored results using the old labels. But you
can use version control to get the old behavior of -mean- (please check
-help version- for more information regarding the version control).

version 15: mean egal_scale [pw=nesw], over(female)

lincom [egal_scale]Male - [egal_scale]Female

Or you may update your code according to Stata 16's behavior
(recommended) and use the new labels in -lincom-. First you can specify
the option -coeflegend- with -mean- to view the labels.

mean egal_scale [pw=nesw], over(female) coeflegend

then you call the estimates using the new labels in -lincom-.

I hope that will help. Please let us know if you have any further
questions.


Best regards,

Mia Lv, Ph.D.
Staff Statistician
StataCorp LLC

[Carlo Lazzaro]

Lauren:
thanks for sharing.

[Samantha Chiu]

Hi All,

Carlo's feedback is correct. I am a student and replicated his lincom solution (specific for ASDS Chapter 5.3 exercises): lincom [email protected]- [email protected].

Thanks! I am going to share this with the class! How do we help rlincom.pdf update their code from the use of brackets?

Samantha

[Lee Payne]

Thank you, Lauren. The coeflegend fixed the lincom issue after the "mean" command, but not after the "proportion" command. You mentioned that mean, proportion, ratio, and total now allow factor-variable notation. Is there a different post-command from coeflegend that works with proportion?

The specific example is below:
proportion domestic_spying [pw=wtss]
lincom [domestic_spying]Oppose - .5

Thanks,
Lee

[John Transue]

Thank you, Lauren. The coeflegend fixed the lincom issue after the "mean" command, but not after the "proportion" command. You mentioned that mean, proportion, ratio, and total now allow factor-variable notation. Is there a different post-command from coeflegend that works with proportion?

The specific example is below:
proportion domestic_spying [pw=wtss]
lincom [domestic_spying]Oppose - .5

Thanks,
Lee

coeflegend does work after proportion and does so in the same way. You need to include that so that you can give lincom the proper information for what you want:

. version 15

. proportion domestic_spying [pw=wtss],coeflegend

Proportion estimation             Number of obs   =      1,362

------------------------------------------------------------------------------
             | Proportion  Legend
-------------+----------------------------------------------------------------
domestic_s~g |
      Oppose |   .5076181  _b[domestic_spying:Oppose]
     Support |   .4923819  _b[domestic_spying:Support]
------------------------------------------------------------------------------

. lincom [domestic_spying]Oppose - .5

 ( 1)  [domestic_spying]Oppose = .5

------------------------------------------------------------------------------
  Proportion |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         (1) |   .0076181   .0151678     0.50   0.616    -.0221367     .037373
------------------------------------------------------------------------------

. 
. version 16

. proportion domestic_spying [pw=wtss],coeflegend

Proportion estimation             Number of obs   =      1,362

---------------------------------------------------------------------------------
                | Proportion  Legend
----------------+----------------------------------------------------------------
domestic_spying |
        Oppose  |   .5076181  _b[0bn.domestic_spying]
       Support  |   .4923819  _b[1.domestic_spying]
---------------------------------------------------------------------------------

. lincom _b[0bn.domestic_spying] - .5

 ( 1)  0bn.domestic_spying = .5

------------------------------------------------------------------------------
  Proportion |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         (1) |   .0076181   .0151678     0.50   0.616    -.0221367     .037373
------------------------------------------------------------------------------

.

[Lee Payne]

Thank you, John. That works perfectly!