–etable– in the Nov. 17 update for v17.0

[John Mullahy]

Users who create post-estimation tables might take note of –etable– in the Nov. 17 update for v17.0. On a quick inspection and test drive it looks rather powerful.

[Zumin Shi]

It is the most useful function. It will be great if Stata can have a predefined table format (e.g. tables in the leading medical journals like JAMA, Lancet) under the etable function.
The current Stata 17 Table function is too complicated. In my opinion, the user-written table1_mc is more user friendly to create Table 1 than the examples in Stata blog.

[Nick Cox]

I think #2 misses the point. A command focused on a specific thing you want to do will usually have the edge of convenience over a more general and complicated command that allows that specific thing to be done with complicated syntax, just as you wouldn't (I guess) try to write your own histogram command using twoway to calculate bin densities (or whatever) and draw each bar as a series of line segments given that a histogram command exists already.

table is in effect a language of its own that allows people to write specific routines to do what they want. If users are not able or willing to invest the effort in learning how to do that for their specific goals, then, fine, their choice, and they will have to await someone else doing it first.

You couldn't use table1_mc (SSC) unless the author had written the best part of a thousand lines of code to make it easier to do things that you want to do. Conversely, "Table 1" rings cheerful bells of recognition with many users but bears no resemblance at all to what many others ever want, which should be fine with everybody.

The analogy with graphics is strong here. Many people can say "All I want is a simple graph" except that so does almost everybody else, and the simple graphs are typically different. and not always as simple as the user implies. The solution has to be on several levels:

1. A very general command that allows programming of most things that someone might want,

2. Specific commands for common needs.

3. Options so that people can tweak cosmetic and other details.

table in Stata 17 is comparable to graph since Stata 8.

[George Hoffman]

is etable working correctly when displaying results from multiple models?
i have run 3 regressions ans stored the results using est sto mod1...
est table mod1 mod2 mod3 shows model coefficients in 3 columns
etable displays results appended and in separate columns, not in horizontally-aligned columns
the .pdf documentation suggests to me that the coefficients should appear aligned in rows

. etable, est(svols rso2cols rso2rols) cstat(_r_b)

------------------------------------------
                         sv   rso2c  rso2r
------------------------------------------
SpO2                   0.598              
Mean Arterial Pressure 0.183              
Intercept              2.650              
SpO2                          0.555       
Mean Arterial Pressure        0.055       
Intercept                    16.713       
SpO2                                 0.420
Mean Arterial Pressure               0.165
Intercept                           32.769
Number of observations  7211   7806   7769
------------------------------------------

. est table svols rso2cols rso2rols

-----------------------------------------------------
    Variable |   svols       rso2cols     rso2rols   
-------------+---------------------------------------
          sa |  .59843087    .55522465    .42013389  
          bp |  .18339393    .05497271     .1650317  
       _cons |  2.6503611    16.712968    32.769461  
-----------------------------------------------------

. 

what am i doing wrong?

[John Mullahy]

I still haven't worked through all the documentation so it's quite possible that there's an option setting that resolves the issue. But here's an additional illustration of the issue identified in #1

est clear
sysuse auto
reg price mpg foreign
est store e1
reg price mpg foreign rep
est store e2
etable, est(e1 e2)

yields

--------------------------------------------
                          price      price  
--------------------------------------------
Mileage (mpg)            -294.196   -292.434
                         (55.692)   (60.227)
Car origin               1767.292   1023.208
                        (700.158)  (866.086)
Repair record 1978                   432.800
                                   (394.705)
Intercept               11905.415  10586.485
                       (1158.634) (1555.745)
Number of observations         74         69
--------------------------------------------

whereas

est clear
sysuse auto
reg price mpg foreign
est store e1
reg headroom mpg foreign rep
est store e2
etable, est(e1 e2)

yields

------------------------------------------
                          price   headroom
------------------------------------------
Mileage (mpg)            -294.196         
                         (55.692)         
Car origin               1767.292         
                        (700.158)         
Intercept               11905.415         
                       (1158.634)         
Mileage (mpg)                       -0.049
                                   (0.018)
Car origin                          -0.495
                                   (0.265)
Repair record 1978                   0.126
                                   (0.121)
Intercept                            3.763
                                   (0.475)
Number of observations         74       69
------------------------------------------

[Leonardo Guizzetti]

For me, I only see the second example of what John shows in #5, and try as I might, I cannot align variables by name.

[George Hoffman]

so i guess the dependent variable has to be the same to yield row alignment of independent variables.
i'm not sure this is a 'feature' or a bug!

[Bjarte Aagnes]

-help Tables Builder- may help to replace the default row specification.

sysuse auto
reg price mpg foreign
est store e1
reg headroom mpg foreign rep78
est store e2
etable, est(e1 e2)

collect layout

Collection: ETable
      Rows: coleq#colname[]#result[_r_b _r_se] result[N]
   Columns: etable_depvar#stars[value]
   Table 1: 15 x 2

collect layout (colname#result[_r_b _r_se] result[N]) (etable_depvar#stars[value])

Collection: ETable
      Rows: colname#result[_r_b _r_se] result[N]
   Columns: etable_depvar#stars[value]
   Table 1: 9 x 2

------------------------------------------
                          price   headroom
------------------------------------------
Mileage (mpg)            -294.196   -0.049
                         (55.692)  (0.018)
Car origin               1767.292   -0.495
                        (700.158)  (0.265)
Repair record 1978                   0.126
                                   (0.121)
Intercept               11905.415    3.763
                       (1158.634)  (0.475)
Number of observations         74       69
------------------------------------------

[John Mullahy]

Thanks Bjarte. This is good to know.

Even so, however, a simple option in the –etable– command could presumably be implemented to accomplish what you did in #8 without the extra steps of referring to the collection.

I imagine that many people wanting to use –etable– to report the results of two or more model estimates would have the same objective as #4, i.e. reporting models that have different LHS variables with either the same or different sets of RHS variables.

For instance, something like

etable, est(e1 e2) alignrhsvars

[Bjarte Aagnes]

Thanks John, maybe the default definition will be changed. The arguments for changing a collect definition via etable options may be made closer to the underlying collect maybe something like:

collect_layout_rows("colname#result[_r_b _r_se]")

BTW the behaviour of etable option name() seems not to meet my expectations, the following will use the row settings of ETable, not the row settings of the etableprefered saved to etableprefered.stjson ("rows": "colname#result[_r_b _r_se] result[N]"):

reg price mpg foreign
est store e1
reg headroom mpg foreign rep78
est store e2
 
etable, est(e1 e2)

collect layout (colname#result[_r_b _r_se] result[N]) (etable_depvar#stars[value])

collect style save etableprefered, replace

collect clear
collect use etableprefered etableprefered.stjson

etable, est(e1 e2) name(etableprefered)

. collect use etableprefered etableprefered.stjson
(collection etableprefered read from file etableprefered.stjson)

. 
. etable, est(e1 e2) name(etableprefered) 

------------------------------------------
                          price   headroom
------------------------------------------
Mileage (mpg)            -294.196         
                         (55.692)         
Car origin               1767.292         
                        (700.158)         
Intercept               11905.415         
                       (1158.634)         
Mileage (mpg)                       -0.049
                                   (0.018)
Car origin                          -0.495
                                   (0.265)
Repair record 1978                   0.126
                                   (0.121)
Intercept                            3.763
                                   (0.475)
Number of observations         74       69
------------------------------------------

[Jeff Pitblado (StataCorp)]

Most commands like logit use the dependent variable as the first
equation name. regress is one of the rare commands that does not
post an equation name to e(b), however collect (and hence
etable) automatically consumes the variable in e(depvar)
as the column equation in this case.

Bjarte shows that you can take control of the layout using
collect and the Tables Builder. However, etable will
always overwrite the layout. So the name() option did not have
the effect Bjarte expected.

etable option eqrecode() allows you to realign equations
by recoding the column equations as you collect results. Options
showeq and novarlabel will help you see the equation names
to recode.

Here is how these options might be used:

sysuse auto
regress price mpg foreign
est store e1
regress headroom mpg foreign rep78
est store e2

* get the results
etable, est(e1 e2)
* peek at the equations
etable, replay showeq novarlabel
* recode the equations
etable, replay eqrecode(price = xb headroom = xb)
* other style changes
etable, replay noshoweq varlabel

* start fresh and recode equations
etable, est(e1 e2) eqrecode(price = xb headroom = xb)

Here is the output from the etable commands:

. * get the results
. etable, est(e1 e2)

------------------------------------------
                          price   headroom
------------------------------------------
Mileage (mpg)            -294.196         
                         (55.692)         
Car origin               1767.292         
                        (700.158)         
Intercept               11905.415         
                       (1158.634)         
Mileage (mpg)                       -0.049
                                   (0.018)
Car origin                          -0.495
                                   (0.265)
Repair record 1978                   0.126
                                   (0.121)
Intercept                            3.763
                                   (0.475)
Number of observations         74       69
------------------------------------------

. * peek at the equations
. etable, replay showeq novarlabel

------------------------------------------
                          price   headroom
------------------------------------------
price                                     
  mpg                    -294.196         
                         (55.692)         
  foreign                1767.292         
                        (700.158)         
  _cons                 11905.415         
                       (1158.634)         
headroom                                  
  mpg                               -0.049
                                   (0.018)
  foreign                           -0.495
                                   (0.265)
  rep78                              0.126
                                   (0.121)
  _cons                              3.763
                                   (0.475)
Number of observations         74       69
------------------------------------------

. * recode the equations
. etable, replay eqrecode(price = xb headroom = xb)

------------------------------------------
                          price   headroom
------------------------------------------
xb                                        
  mpg                    -294.196   -0.049
                         (55.692)  (0.018)
  foreign                1767.292   -0.495
                        (700.158)  (0.265)
  rep78                              0.126
                                   (0.121)
  _cons                 11905.415    3.763
                       (1158.634)  (0.475)
Number of observations         74       69
------------------------------------------

. * other style changes
. etable, replay noshoweq varlabel

------------------------------------------
                          price   headroom
------------------------------------------
Mileage (mpg)            -294.196   -0.049
                         (55.692)  (0.018)
Car origin               1767.292   -0.495
                        (700.158)  (0.265)
Repair record 1978                   0.126
                                   (0.121)
Intercept               11905.415    3.763
                       (1158.634)  (0.475)
Number of observations         74       69
------------------------------------------

. 
. * start fresh and recode equations
. etable, est(e1 e2) eqrecode(price = xb headroom = xb)

------------------------------------------
                          price   headroom
------------------------------------------
Mileage (mpg)            -294.196   -0.049
                         (55.692)  (0.018)
Car origin               1767.292   -0.495
                        (700.158)  (0.265)
Repair record 1978                   0.126
                                   (0.121)
Intercept               11905.415    3.763
                       (1158.634)  (0.475)
Number of observations         74       69
------------------------------------------

[John Mullahy]

Thanks Jeff. This is really helpful.

[Antonio Arruebo]

Hi there! Extremely helpful thread, thanks for helping the community out so effectively. Now that I have been to horizontally-align the two models' results by using #11's code, how can I include the log pseudo likelihood indicator and change the column's heading?

Thanks in advance,
Antonio