Column and Row labels go away upon sorting

[Jared Greathouse]

EDIT: I use gcollapse (ssc inst gtools, replace), but regular collapse works fine.




When I sort a matrix like this

* Example generated by -dataex-. For more info, type help dataex
clear
input long code float(premse postmse ratio rmspe_rank p)
11  .05515488  2.771367    50.247 1 .16666667
12  .14310792  .2025988 1.4157065 6         1
26 .012931863  .5920703  45.78384 3        .5
27  .02456784 1.1890676  48.39935 2  .3333333
35 .029555114  .3298432 11.160275 5  .8333333
51   .0257518   .880294 34.183784 4  .6666667
end
label values code code
label def code 11 "Rondônia", modify
label def code 12 "Acre", modify
label def code 26 "Pernambuco", modify
label def code 27 "Alagoas", modify
label def code 35 "São Paulo", modify
label def code 51 "Mato Grosso", modify

gcollapse (firstnm) premse postmse ratio rmspe_rank p, by(code)

decode code, g(panel)

mkmat premse postmse ratio rmspe_rank p, rownames(panel) mat(errmat)

matname errmat "Pre-MSE" "Post-MSE" "Ratio" "Rank" "p", columns(1..5) explicit

mata : st_matrix("errmat", sort(st_matrix("errmat"), 4))


mat l errmat

it robs me of my precious row and column labels.

. mat l errmat

errmat[6,5]
               Pre-MSE   Post-MSE      Ratio       Rank          p
   Rondônia  .05515488  2.7713671  50.247002          1  .16666667
       Acre  .14310792   .2025988  1.4157065          6          1
 Pernambuco  .01293186  .59207028   45.78384          3         .5
    Alagoas  .02456784  1.1890676  48.399349          2  .33333331
  São_Paulo  .02955511  .32984319  11.160275          5  .83333331
Mato_Grosso   .0257518  .88029402  34.183784          4  .66666669

. mata : st_matrix("errmat", sort(st_matrix("errmat"), 4))

.
.
. mat l errmat

errmat[6,5]
           c1         c2         c3         c4         c5
r1  .05515488  2.7713671  50.247002          1  .16666667
r2  .02456784  1.1890676  48.399349          2  .33333331
r3  .01293186  .59207028   45.78384          3         .5
r4   .0257518  .88029402  34.183784          4  .66666669
r5  .02955511  .32984319  11.160275          5  .83333331
r6  .14310792   .2025988  1.4157065          6          1

Why might this be?

[daniel klein]

The technical answer to "why" is that Mata matrices do not have row or column names. Mata can obtain and attach row and column names from and to Stata matrices but that is a separate step from obtaining the matrix contents, i.e., values. The conceptual answer to "why" is for StataCorp to answer.

Anyway, I show a possible work around here. I refer to the wrong function name in the final code-snippet, which you should obviously not do.