fsum order of the statistics

[Gregoire Urvoy]

Hello,

Does anyone knows if it is possible to change the order of the statistics in a fsum table ?

For example I have fsum X, stats (mean sd n).

But I have this table:

Variable______________________| N___ Mean___ SD
------------------------------------------------+---------------------------
X __________________________| 148 7. 96 1.06

And I would like something like:
Variable______________________| Mean___ SD___ N
------------------------------------------------+---------------------------
X __________________________| 7. 96 1.06 148

Thank you very much,

Grégoire

[Nick Cox]

tabstat lets you specify an order for the statistics shown.

.. sysuse auto, clear
(1978 Automobile Data)

. tabstat mpg, s(mean sd N) by(rep78)

Summary for variables: mpg
     by categories of: rep78 (Repair Record 1978)

   rep78 |      mean        sd         N
---------+------------------------------
       1 |        21  4.242641         2
       2 |    19.125  3.758324         8
       3 |  19.43333  4.141325        30
       4 |  21.66667   4.93487        18
       5 |  27.36364  8.732385        11
---------+------------------------------
   Total |  21.28986  5.866408        69
----------------------------------------

[Gregoire Urvoy]

I know but I need to use fsum to have the labels in my table.

Unfortunately, tabstat puts the name of the variable and not the label.

[Nick Cox]

What you want is programmable by yourself. e.g.

sysuse auto, clear  

gen which = ""
gen mean = .
gen sd = .
gen N = .

local i = 0
quietly foreach v in mpg weight turn trunk {
    local ++i
    replace which = "`: var label `v''" in `i'
    replace which = "`v'" in `i' if missing(which[`i'])
    su `v'
    replace mean = r(mean) in `i'
    replace sd = r(sd) in `i'
    replace N = r(N) in `i'
}

char which[varname] " "
replace which = trim(which)
format mean sd %2.1f
l which-N in 1/`i', noobs subvarname  

  +---------------------------------------------+
  |                           mean      sd    N |
  |---------------------------------------------|
  |         Mileage (mpg)     21.3     5.8   74 |
  |         Weight (lbs.)   3019.5   777.2   74 |
  |     Turn Circle (ft.)     39.6     4.4   74 |
  | Trunk space (cu. ft.)     13.8     4.3   74 |
  +---------------------------------------------+


tabdisp which in 1/4, c(mean sd N)

----------------------------------------------------------
                which |       mean          sd           N
----------------------+-----------------------------------
        Mileage (mpg) |       21.3         5.8          74
Trunk space (cu. ft.) |       13.8         4.3          74
    Turn Circle (ft.) |       39.6         4.4          74
        Weight (lbs.) |     3019.5       777.2          74
----------------------------------------------------------

[Gregoire Urvoy]

Thank you very much.

I did it and it works perfectly well.

I have a last problem: the labels are ordered alphabetically. Do you have any suggestions ?

I could use the first table but I have this:

[Gregoire Urvoy]

And I would like to have something like this:

Thank you so much for you help.

[Nick Cox]

You don't give which code you are using, but it seems the tabdisp code. Use the list code instead.

[Gregoire Urvoy]

My problem with the list code is the lines in the table.

My problem with the tabdisp code is the fact that the labels are ordered alphabetically.

Do you see what I mean ?

Thank you

[Nick Cox]

Quite. I think you need to write your own program or just get to work editing the output.

[Gregoire Urvoy]

I am just a beginner on stata. So it looks a bit complicated for me. No worries, I will find something else.

Thank you for your time.

[Stephen Jenkins]

Ben Jann's estout and its wrapper program esttab have good facilities for formatting summary statistics (as well as regression estimates), and you may find them useful: seehttp://repec.org/bocode/e/estout/est...tml#estpost101. Download from SSC: ssc describe estout