topic Re: Lining up variables when exporting data to Excel from PROC IML in SAS Procedures

Lining up variables when exporting data to Excel from PROC IML

dvtarasov — Thu, 12 Nov 2015 15:36:33 GMT

I have a dataset in which three variables (let's call them X, Y, and Z) are to be written into matrices in PROC IML and have some calculations performed upon them. The result of the calculations I need to export to an Excel file. Another column in the original dataset (let's call it N) will not be used in any calculations, but I need it written to the Excel spreadsheet alongside the output of the calculation.

How do I ensure that the values in N line up with the corresponding results in the output? I suspect that a loop of some kind is needed (for every record in N, (write X, Y, and Z to matrices, perform the calculations, write N to Excel, write results to Excel)). Am I right about the loop? Is it possible to have several different steps within one loop? And what is the exact syntax for this?

Re: Lining up variables when exporting data to Excel from PROC IML

PaigeMiller — Thu, 12 Nov 2015 15:47:23 GMT

If PROC IML is not changing the order of the data, you would write the IML results into a SAS data set, and then merge with the original data set that contains N.

If IML is changing the order of the data, you'd need to include N in your IML matrices and sort/re-arrange N the same as you are doing with the rest of the data.

Re: Lining up variables when exporting data to Excel from PROC IML

dvtarasov — Thu, 12 Nov 2015 16:00:36 GMT

Suppose I now have a matrix with two columns (one read from N, the other from X), and I need to multiply the matrix elementwise by the matrix that was read from Y. What is the syntax for leaving the column in the first matrix that came from N unchanged?

Re: Lining up variables when exporting data to Excel from PROC IML

PaigeMiller — Thu, 12 Nov 2015 16:06:36 GMT

Assuming matrix X has variable X in column 1 and variable N in column 2

newcolumn=x[:,1]#y;

But it sounds like there's really no reason to put X and N into the same matrix in IML.