Trouble with tables

[Niels Henrik Bruun]

Hi
I'm trying to use Stata's Tables.
But somehow I am getting the standard deviation twice:

. sysuse auto
(1978 automobile data)

. table rep78, stat(mean price) stat(sd price) command(_r_ci:mean price)

------------------------------------------------------------------------------------------------
                   |      Mean   Standard deviation                   mean price                
                   |      Mean   Standard deviation   Standard deviation           95% CI       
-------------------+----------------------------------------------------------------------------
Repair record 1978 |                                                                            
  1                |    4564.5             522.5519             522.5519   -130.4427    9259.443
  2                |  5967.625             3579.357             3579.357    2975.208    8960.042
  3                |  6429.233              3525.14              3525.14    5112.924    7745.542
  4                |    6071.5             1709.608             1709.608    5221.332    6921.668
  5                |      5913             2615.763             2615.763    4155.707    7670.293
  Total            |  6146.043              2912.44              2912.44    5446.399    6845.688
------------------------------------------------------------------------------------------------

What am I doing wrong?

[Simon Pfaff]

Hallo Niels,

actually, I am not sure why standard deviation is printed twice. However, you could use "collect layout" for arranging the table (in Stata 17 or 18).

sysuse auto, clear

table rep78, stat(mean price) stat(sd price) command(_r_ci:mean price)

collect layout (rep78)(result[mean] result[sd]  result[_r_ci])

[Jeff Pitblado (StataCorp)]

To see what is going on with your layout, change the header labeling
to show dimension level values.

* get a report of the dimensions used in the layout
collect layout
* show the level labels of the column dimensions
collect label list statcmd, all
collect label list result, all
* show the level values in the column headers
collect style header statcmd result, level(value)
* replay layout
collect layout

This is what I get with your example.

. * get a report of the dimensions used in the layout
. collect layout

Collection: Table
      Rows: rep78
   Columns: statcmd#result
   Table 1: 7 x 4

------------------------------------------------------------------------------------------------
                   |      Mean   Standard deviation                   mean price
                   |      Mean   Standard deviation   Standard deviation           95% CI
-------------------+----------------------------------------------------------------------------
Repair record 1978 |
  1                |    4564.5             522.5519             522.5519   -130.4427    9259.443
  2                |  5967.625             3579.357             3579.357    2975.208    8960.042
  3                |  6429.233              3525.14              3525.14    5112.924    7745.542
  4                |    6071.5             1709.608             1709.608    5221.332    6921.668
  5                |      5913             2615.763             2615.763    4155.707    7670.293
  Total            |  6146.043              2912.44              2912.44    5446.399    6845.688
------------------------------------------------------------------------------------------------

. * show the level labels of the column dimensions
. collect label list statcmd, all

  Collection: Table
   Dimension: statcmd
       Label: Statistic/command option index
Level labels:
           1  Mean
           2  Standard deviation
           3  mean price

. collect label list result, all

  Collection: Table
   Dimension: result
       Label: Result
Level labels:
           N  Number of observations
      N_over  Number of subpopulations
          _N  Nonmissing observations
        _r_b  Coefficient
       _r_ci  __LEVEL__% CI
       _r_df  df
       _r_lb  __LEVEL__% lower bound
        _r_p  p-value
       _r_se  Std. error
       _r_ub  __LEVEL__% upper bound
        _r_z  t
    _r_z_abs  |t|
         cmd  Command
     cmdline  Command line as typed
        df_r  Sample DF
       error
   estat_cmd  Program used to implement estat
        k_eq  Number of equations
marginsnotok  Predictions disallowed by margins
        mean  Mean
  properties  Command properties
        rank  Rank of VCE
          sd  Standard deviation
       title  Title of output
     varlist  Variables
         vce  SE method

. * show the level values in the column headers
. collect style header statcmd result, level(value)

. * replay layout
. collect layout

Collection: Table
      Rows: rep78
   Columns: statcmd#result
   Table 1: 7 x 4

----------------------------------------------------------------------------
                   |         1          2                   3
                   |      mean         sd         sd           _r_ci
-------------------+--------------------------------------------------------
Repair record 1978 |
  1                |    4564.5   522.5519   522.5519   -130.4427    9259.443
  2                |  5967.625   3579.357   3579.357    2975.208    8960.042
  3                |  6429.233    3525.14    3525.14    5112.924    7745.542
  4                |    6071.5   1709.608   1709.608    5221.332    6921.668
  5                |      5913   2615.763   2615.763    4155.707    7670.293
  Total            |  6146.043    2912.44    2912.44    5446.399    6845.688
----------------------------------------------------------------------------

From the labeling information we see that level 2 of dimension
statcmd has label "Standard deviation", as does level sd
of dimension result.

Note that result sd shows up from option stat() for
statcmd=2, but also shows up from matrix e(sd) from the
mean command for statcmd=3. This is the source of your
duplicate column of standard deviations.

. * -table- left behind the results of -mean price- for the sample where
. * rep78=5
. matrix list e(sd)

symmetric e(sd)[1,1]
        price
y1  2615.7628

Here is the table, restoring the result level labels, and
assuming you want just the statistic names in the column header.

. * restore the header styles
. collect style header statcmd result, level(label)

. collect style header statcmd, level(hide)

.
. * a more selective layout
. collect layout (rep78) (statcmd[1 2]#result statcmd[3]#result[_r_ci])

Collection: Table
      Rows: rep78
   Columns: statcmd[1 2]#result statcmd[3]#result[_r_ci]
   Table 1: 7 x 3

---------------------------------------------------------------------------
                   |      Mean   Standard deviation           95% CI      
-------------------+-------------------------------------------------------
Repair record 1978 |                                                      
  1                |    4564.5             522.5519   -130.4427    9259.443
  2                |  5967.625             3579.357    2975.208    8960.042
  3                |  6429.233              3525.14    5112.924    7745.542
  4                |    6071.5             1709.608    5221.332    6921.668
  5                |      5913             2615.763    4155.707    7670.293
  Total            |  6146.043              2912.44    5446.399    6845.688
---------------------------------------------------------------------------

[Niels Henrik Bruun]

Simon Pfaff Thank you for your suggestion
Jeff Pitblado (StataCorp) I have a bad habit of not reading the manual before too late.
So what is happening is that the information, I want to show is already in the ereturn and I could just as well have done:

. qui table (rep78), command(e(N) e(b) _r_b e(sd) _r_ci:mean price) nformat(%9.1f)

. collect style cell result[N], nformat(%9.0f)

. collect preview

--------------------------------------------------------------------------------------------------------------
                   |  Number of observations   Coefficient   Standard deviation estimates         95% CI      
                   |                                 Price                          Price          Price      
-------------------+------------------------------------------------------------------------------------------
Repair record 1978 |                                                                                          
  1                |                       2        4564.5                          522.6   -130.4      9259.4
  2                |                       8        5967.6                         3579.4   2975.2      8960.0
  3                |                      30        6429.2                         3525.1   5112.9      7745.5
  4                |                      18        6071.5                         1709.6   5221.3      6921.7
  5                |                      11        5913.0                         2615.8   4155.7      7670.3
  Total            |                      69        6146.0                         2912.4   5446.4      6845.7
--------------------------------------------------------------------------------------------------------------

PS Tables is a fantastic addition to Stata