topic Re: Matrix reshaping in SAS/IML Software and Matrix Computations

Matrix reshaping

xtc283x — Fri, 09 Oct 2015 10:53:27 GMT

I have a matrix, x, that is numeric and 49X47. The second row contains all missing values. I would like to reshape it to a square matrix that is 47x47 by eliminating the first and second rows.

The SHAPE function isn't working:

x2=shape(x,3:49,47);

I'm getting execution errors telling me that the "argument should be a scalar."

How do I resolve this?

Thank you.

Re: Matrix reshaping

Rick_SAS — Fri, 09 Oct 2015 12:06:44 GMT

The documentation for the SHAPE function specifies that the second argument should be the number of columns in the result matrix. You are sending in a vector, which is why you are getting an error.

First use the subscript operator to extract the nonmissing rows, then reshape:

x2 = shape( x[3:49, ], 47 );

Re: Matrix reshaping

IanWakeling — Fri, 09 Oct 2015 12:35:12 GMT

I think the subscript operator alone is suficient here. The shape function does nothing as the reduced matrix already has 47 rows and 47 columns.

Re: Matrix reshaping

xtc283x — Mon, 12 Oct 2015 12:08:29 GMT

That specification of the Shape function worked. When I print the resulting matrix within IML, it is square. However, when I create output containing that matrix, the output resolves to a vector:

Vec2 = the resulting square matrix

create Outputx var {Vec2};
append;
close Outputx;

run;*end Proc IML;

Outputx contains a single column with nxn rows.

Why does Outputx not contain the square matrix Vec2?

Thank you.

Re: Matrix reshaping

Rick_SAS — Mon, 12 Oct 2015 12:15:26 GMT

Why does Outputx not contain the square matrix Vec2?

Because you are using the VAR clause in the CREATE statement. Use the FROM clause to write a matrix, as discussed in the article "Writing data from a matrix to a SAS data set."