Sergiy, thank you. Your routine does fulfill part of the requirement
in allowing data to be sorted in label alphabetical order. However,
the other requirement of alphabetical ordering in tables is not met.
The merge method is surprisingly quick compared to the recode. I ran
my routine and yours with -profiler on- and it was faster despite the
preserve,sort and merge steps. I will consider refactoring my code in
light of this although I would still like to be able to do the
recoding within mata, or save the sorted matrix with old and new
values to a .dta file capable of being merged. More poring over the
manual is in order, obviously.
If you are interested - here are the profiler reports:
. set memory 500m
. sysuse nlsw88.dta, clear
. expand 1000
. profiler on
. labvalsort industry
. profiler report
labvalsort
1 0.218 labvalsort
recode
1 24.437 recode
24 0.000 GetEl
12 0.000 CheckParens
24.437 Total
unab
1 0.000 unab
label
5 0.000 label
recast
1 1.344 recast
move
1 0.297 move
lab
1 0.000 lab
Overall total count = 47
Overall total time = 26.296 (sec)
. profiler clear
labels_ABC industry
. profiler report
labels_ABC
1 11.734 labels_ABC
uselabel
1 0.000 uselabel
1 0.000 Uselabel_u
0.000 Total
label
3 0.000 label
merge
1 1.797 merge
unab
1 0.000 unab
Overall total count = 8
Overall total time = 13.531 (sec)
Clearly recode may not be the command to use.
DCE
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
