topic Re: How to find the location of miss value in a matrix? in SAS/IML Software and Matrix Computations

How to find the location of miss value in a matrix?

ShufeGuoding — Fri, 19 Dec 2014 07:15:34 GMT

I want to replace missing value in a matrix A with zero and first need to find their locations in the matrix. Since function loc() treats the missing values as zero, how can I use the function to find their locations in A?

Re: How to find the location of miss value in a matrix?

Rick_SAS — Fri, 19 Dec 2014 11:19:15 GMT

proc iml;

A = {. 2 3, 4 . 6};

idx = loc(A=.);

A[idx] = 0;

Re: How to find the location of miss value in a matrix?

ShufeGuoding — Sun, 21 Dec 2014 00:28:18 GMT

Because missing values are treated as zero, Your code seems not work properly. The problem is the same as that I met.

Re: How to find the location of miss value in a matrix?

Reeza — Sun, 21 Dec 2014 00:37:56 GMT

Works fine for me, idx identifies the missing as 1, 5. Next line replaces them to be 0's.

proc iml;

A = {. 2 3, 4 . 6};

idx = loc(A=.);

print idx;

A[idx] = 0;

print a;

OUTPUT:

The SAS System

0	2	3
4	0	6

Re: How to find the location of miss value in a matrix?

ShufeGuoding — Sun, 21 Dec 2014 08:32:57 GMT

I get it. Thanks a lot!

Steve

Re: How to find the location of miss value in a matrix?

Rick_SAS — Sun, 21 Dec 2014 18:12:04 GMT

As far as I know, missing values are not "treated as zero" anywhere in IML. You might be thinking of the behavior of an IF/THEN-ELSE statement, where a missing value in the IF conditional results in the condition evaluating to FALSE.

Re: How to find the location of miss value in a matrix?

ShufeGuoding — Mon, 22 Dec 2014 00:23:58 GMT

I was misleaded by the statements for Loc function in Usg/IML12.1 that read:

The loc funciton finds nonzero elements of a matix. .... . Missing values treated as zeros.:smileyconfused:

Re: How to find the location of miss value in a matrix?

Rick_SAS — Mon, 22 Dec 2014 12:40:09 GMT

I see. I interpret that phrase to mean that

y = loc(x);

will produce a matrix y such that y[i,j]=0 whenever x[i,j] is zero or missing, but I see how that phrase might be confusing. To make sure you always get the correct answer, I recommend using a logical expression as the argument to the LOC function. For example

y = loc(x^=0);

y = loc(x>0);

etc