changing elements in a matrix according to [row,column] coordinates in another matrix

[Jose Uribe]

Dear statalist,

I have a matrix "X" of zeroes of dimensions (r by k).

I have a second matrix "C" of dimensions (n by 2), where each row denotes a position in "X" where I want to replace a zero for a 1.

For example, the first three rows in "C" are:

1 3
4 24
6 39

I want to replace into ones the zeroes in positions [1,3], [4,2] and [6,39] in "X" .

I am using Stata version 13.1.
Thank you very much
Jose Uribe

[Phil Schumm]

for (i=1; i<=rows(C); i++) {
    X[C[i,1],C[i,2]] = 1
}

[Jose Uribe]

Thanks Phil, that is exactly what I needed!

[Christophe Kolodziejczyk]

It should also work and avoids the loop

k = max( (rows(C),cols(C)))
X[C[,1],C[,2]] = J(k,k,1)

Best
Christophe

[Christophe Kolodziejczyk]

Sorry for misleading you with my last post, since it does not give the intended result. So forget it.

[Christophe Kolodziejczyk]

ok last try :-).

k = max( (rows(C),cols(C)))
X[C[,1],C[,2]] = I(k)

[Jose Uribe]

Christophe, your code is appealing but I did not get the desired answer (as I did with the loop).
I have another question, intimately related to the first:
After some operations on X, I now have matrix Y, also of dimensions (r by k).

I want to "extract" only those entries in Y indexed by the coordinates in matrix C dimension (n by 2) and append them to matrix C to create matrix W dimension (n by 3) .

For example, the first three rows in "C" dimension (n by 2) are:

1 3
4 24
6 39

Let's say the values in Y indexed by [1,3], [4,2] and [6,39] are 0, 5 and 10.

The resulting matrix W should be

1 3 0
4 24 5
6 39 10



Thank you very much
Jose Uribe

[Phil Schumm]

Christophe, your code is appealing but I did not get the desired answer (as I did with the loop).

Christophe's approach is neat, but it depends on assumptions about C (i.e., that it specifies at most one cell per row of X) that may not be true. I presume that's why it didn't yield the correct result in your case.

All is not lost, however; we can use Christophe's suggested use of subscripts to solve your second question (and nicely so):

W = C, diagonal(Y[C[,1],C[,2]])

Note that while it's perfectly acceptable to ask narrow questions like this, stitching together a bunch of code fragments is unlikely to result in an efficient overall solution. That may well not matter in your case; indeed, with the speed of modern computers, it often doesn't. However, you might also consider describing more of the problem you are trying to solve, which might permit people to suggest a more efficient (and/or simpler) overall approach. Also, if you're going to work in Mata (and you definitely should!), spending an afternoon going through selected sections of the Mata Reference Manual is an invaluable investment.

[Jose Uribe]

Thanks very much Phil. The code works like a charm. I have and will continue to study the Mata Reference Manual.
Best,
Jose