How to determine the column-number of a matrix element?

[John Whymer]

I have a matrix like:

1
+----------------+
1 | -.4755687238 |
2 | -7.845191301 |
3 | -7.976995867 |
4 | -7.968100143 |
5 | -8.008870078 |
6 | -7.995388117 |
7 | -7.982097707 |
8 | -7.985624132 |
9 | -7.972667072 |
10 | -7.95900065 |
11 | -7.966202475 |
12 | -7.980775105 |
13 | -7.967044048 |
+----------------+

I have determined the smallest element of this matrix with

colmin(b[.,1])

It is equal to -8.008870078. Now I want MATA to determine the column-number of -8.008870078, which is 5. How can I do this?

Sorry, this might be a typical beginner question! But I can't find a solution to my problem in the manual. Thanks a lot for any help!

John

[Scott Merryman]

Here is one way:

clear*
input a double b
1  -.4755687238 
2  -7.845191301 
3  -7.976995867 
4  -7.968100143 
5  -8.008870078 
6  -7.995388117 
7  -7.982097707 
8  -7.985624132 
9  -7.972667072 
10  -7.95900065 
11  -7.966202475 
12  -7.980775105 
13  -7.967044048 
end 
putmata m = b
mata
for (i= 1; i <= rows(m); i++){
    if( colmin(m)==m[i]) {
        ColNum = i
    }
}
ColNum
end

[John Mullahy]

John: Also perhaps take a look at Mata's minindex function and see if that might suit your objectives: help mf_minindex.

[Niels Henrik Bruun]

To find a value in a vector I often use:

:         m = (-.4755687238, -7.845191301, 7.976995867, -7.968100143, -8.008870078, 
>                 -7.995388117, -7.982097707, -7.985624132, -7.972667072, -7.95900065, 
>                 -7.966202475, -7.980775105, -7.967044048)'

:         select(1::rows(m), m:==colmin(m))
  5

[John Whymer]

Thanks a lot to you all!

I got Scotts approach working to select the AIC lags from varsoc and transfer it to a dfuller test:

varsoc ln_CPI_FRA_sa_shw1
mata
b = st_matrix("r(stats)")
m = b[.,7]

for (i= 1; i <= rows(m); i++){
    if( colmin(m)==m[i]) {
        ColNum = i
    }
}
ColNum
st_numscalar("optilag", ColNum)
end
local xlags = optilag
dfuller ln_CPI_FRA_sa_shw1 if month >= 0 & month <= 156, lags(`xlags') trend

And I got Niels approach working for the same task:

varsoc ln_CPI_FRA_sa_shw1
mata
b = st_matrix("r(stats)")
m = b[.,7]
ColNum = select(1::rows(m), m:==colmin(m))
st_numscalar("optilag", ColNum)
end
local xlags = optilag
dfuller ln_CPI_FRA_sa_shw1 if month >= 0 & month <= 156, lags(`xlags') trend

John's approach did somehow not work, but most likely I did not try hard enough, since I have already a workable solution... Thanks a lot!

[Christophe Kolodziejczyk]

you don't say which version of Stata you're using, but since version 14 there is a selectindex() function which can used in this case instead of select().

 
 ColNum = selectindex(m:==colmin(m))