[Chen Samulsion]

Dear Stephen Jenkins, thank you very much for your elaboration.

[Ben Jann]

Ben's -pshare- with the step option provides results matching those of -sumdist- (and -svylorenz-).

... as long as there are no ties. Option -step- only omits interpolation of flat regions, ties will still be broken.

[Ben Jann]

Follow #13: I found that -sumdist- & -pshare- had crossed at least 6 years ago https://www.statalist.org/forums/for...ist-and-pshare
BTW, could the -sumdist- add another column to display mean of variable in each decile group? Thus we can avoid to execute xtile and tabstat to get the mean of each decile group. Although with the option qgp(newvarname), we can create a new variable (say qgpvar) that identifies the quantile group membership of each observation, and run tabstat variable, by(qgpvar) to get what I want.

You could use option -average- in pshare. Example:

. sysuse nlsw88
. pshare wage, n(10) average

Percentile shares (average)              Number of obs = 2,246

--------------------------------------------------------------
        wage | Coefficient  Std. err.     [95% conf. interval]
-------------+------------------------------------------------
        0-10 |   2.661352   .0466623      2.569846    2.752858
       10-20 |   3.566543   .0581448       3.45252    3.680567
       20-30 |   4.306469   .0591152      4.190542    4.422395
       30-40 |   5.042257     .07195      4.901162    5.183353
       40-50 |   5.858092   .0849103      5.691581    6.024603
       50-60 |   6.783161   .1042711      6.578683    6.987639
       60-70 |   7.957788   .1317386      7.699445     8.21613
       70-80 |   9.500946   .1409769      9.224487    9.777405
       80-90 |    11.3826   .1721119      11.04509    11.72012
      90-100 |   20.61028   .7789552      19.08273    22.13783
--------------------------------------------------------------

This gives you the average of the running variable within each decile group. If you are interested in the averages of different variables, you can use option -pvar()- to specify the running variable. Example (note that the estimation sample changes because -tenure- has additional missing values; for reasons of consistency you may want to exclude these observations also from the above command):

. pshare tenure, n(10) average pvar(wage)

Percentile shares (average)              Number of obs = 2,231

--------------------------------------------------------------
      tenure | Coefficient  Std. err.     [95% conf. interval]
-------------+------------------------------------------------
        0-10 |   2.746612   .2253933       2.30461    3.188615
       10-20 |   3.397478   .2678284      2.872259    3.922697
       20-30 |   4.008159     .29548      3.428714    4.587603
       30-40 |   5.321381     .33153      4.671241     5.97152
       40-50 |   5.188225   .3270064      4.546957    5.829494
       50-60 |   7.141261   .3634414      6.428542     7.85398
       60-70 |   7.397206   .3920051      6.628473    8.165939
       70-80 |   8.358247   .4231387       7.52846    9.188034
       80-90 |   8.538137   .4036212      7.746624    9.329649
      90-100 |   7.681794   .4063035      6.885022    8.478567
--------------------------------------------------------------
(percentile groups with respect to wage)

This also works with multiple outcome variables:

. pshare tenure ttl_exp, n(10) average pvar(wage)

Percentile shares (average)              Number of obs = 2,231

--------------------------------------------------------------
             | Coefficient  Std. err.     [95% conf. interval]
-------------+------------------------------------------------
tenure       |
        0-10 |   2.746612   .2253933       2.30461    3.188615
       10-20 |   3.397478   .2678284      2.872259    3.922697
       20-30 |   4.008159     .29548      3.428714    4.587603
       30-40 |   5.321381     .33153      4.671241     5.97152
       40-50 |   5.188225   .3270064      4.546957    5.829494
       50-60 |   7.141261   .3634414      6.428542     7.85398
       60-70 |   7.397206   .3920051      6.628473    8.165939
       70-80 |   8.358247   .4231387       7.52846    9.188034
       80-90 |   8.538137   .4036212      7.746624    9.329649
      90-100 |   7.681794   .4063035      6.885022    8.478567
-------------+------------------------------------------------
ttl_exp      |
        0-10 |   8.975308   .3307737      8.326651    9.623964
       10-20 |   10.37355   .2991184      9.786972    10.96013
       20-30 |   10.63966   .3020635      10.04731    11.23202
       30-40 |   12.17331   .2859634      11.61253     12.7341
       40-50 |   11.77993   .2975009      11.19652    12.36334
       50-60 |   13.52641   .2726825      12.99167    14.06115
       60-70 |   13.81569   .2976969      13.23189    14.39948
       70-80 |   14.43039   .2667726      13.90724    14.95354
       80-90 |   14.98143   .2448512      14.50127    15.46159
      90-100 |   14.85282   .2347553      14.39246    15.31318
--------------------------------------------------------------
(percentile groups with respect to wage)

Equivalent commands in -dstat- are:

. dstat share wage, n(10) average
. dstat share tenure, n(10) average by(wage)
. dstat share tenure ttl_exp, n(10) average by(wage)

ben

[Chen Samulsion]

Thank you very much professor ben!