Minimum or maximum from a set contained in global macro

[Francois Durant]

Hi,

I would like to create a local macro that would contain only the maximum value of a set contained in a global macro:

global M_per 6 12 24 36 72 120

* I would like to input the max of $M_per, which is 120 in the local variable max_M_Per 
local max_M_Per 120

Can someone help?

[Joseph Coveney]

. global M_per 6 12 24 36 72 120

. local max_M_per = max(`: subinstr global M_per " " ", ", all')

. di "`max_M_per'"
120

I'm not sure what you're trying to do, but it sure sounds as if you'd benefit from learning Mata.

[Joro Kolev]

Same thing that Joseph does, but in two steps, and with a clarification: the -max- function does exactly what you want, however it operates on comma delimited lists, and you have a space delimited list in your macro:

. local M_per 6 12 24 36 72 120

. local M_per : subinstr local M_per " " ",", all

. dis "`M_per'"
6,12,24,36,72,120

. local max_M_per = max(`M_per')


.  dis "`max_M_per'"
120

[Nick Cox]

As a small exercise in making the simple complicated

. local M_per 6 12 120 24 36 72

. mata: st_numscalar("max", max(strtoreal(tokens(st_local("M_per"))')))

. scalar li max
       max =        120

[Francois Durant]

Is Nick's solution more general or robust than Joseph's (Nick published after Joseph)? Sorry to ask, I'm not proficient in Mata. And thanks to all!

[Nick Cox]

No; it is not more general or robust. I was riffing on the theme by Joseph Coveney, running that Mata can be useful too.

The one complication I can think of is missing values, but the max() function in Stata and the max() function in Mata both ignore missing values to the extent possible.

[Joro Kolev]

Francois can come up with other complications, Nick.
Here is one complication he came up with on this thread
https://www.statalist.org/forums/for...bal-macro-list
presence of comment in the macro.

I do not know which notion of "robustness" Francois has in mind, but the two approaches respond differently in the face of this complication:

. local M_per 6 12 24 36 72 120 /*240*/

. local M_per : subinstr local M_per " " ",", all

. dis "`M_per'"
6,12,24,36,72,120,/*240*/

. local max_M_per = max(`M_per')
[/] not found
r(111);

. local M_per 6 12 24 36 72 120 /*240*/

. mata: st_numscalar("max", max(strtoreal(tokens(st_local("M_per"))')))

. scalar li max
       max =        120


.

No; it is not more general or robust. I was riffing on the theme by Joseph Coveney, running that Mata can be useful too.

The one complication I can think of is missing values, but the max() function in Stata and the max() function in Mata both ignore missing values to the extent possible.

[Joseph Coveney]

I do not know which notion of "robustness" Francois has in mind, but the two approaches respond differently in the face of this complication

Again, the difference is observed only when executing the code from the command line (see below).

The other notion of robustness that maybe you're alluding to can be summarized as: given that such block commenting is not supposed to be allowed at the command line, and so is either a syntax error in that context or else it's to be taken literally, the first approach is in this sense more robust in that it gives rise to an error.

.ÿ
.ÿversionÿ16.1

.ÿ
.ÿclearÿ*

.ÿ
.ÿlocalÿM_perÿ6ÿ12ÿ24ÿ36ÿ72ÿ120ÿ/*240*/

.ÿlocalÿM_perÿ:ÿsubinstrÿlocalÿM_perÿ"ÿ"ÿ",",ÿall

.ÿdisÿ"`M_per'"
6,12,24,36,72,120

.ÿlocalÿmax_M_perÿ=ÿmax(`M_per')

.ÿdisÿ"`max_M_per'"
120

.ÿlocalÿM_perÿ6ÿ12ÿ24ÿ36ÿ72ÿ120ÿ/*240*/

.ÿmata:ÿst_numscalar("max",ÿmax(strtoreal(tokens(st_local("M_per"))')))

.ÿscalarÿliÿmax
ÿÿÿÿÿÿÿmaxÿ=ÿÿÿÿÿÿÿÿ120

.ÿ
.ÿexit

endÿofÿdo-file


.