topic Re: Alternative to proc iml? in SAS/IML Software and Matrix Computations

Alternative to proc iml?

Sternchen — Mon, 14 Apr 2014 13:17:36 GMT

Hello everyone,

I have just started with SAS and am already stuck; I really hope that someone can help me out *please*

I have been provided a macro to create a new variable within my dataset, but it contains "proc iml" which, apparently,

my SAS Software does not have (can not use?...)

So my question is, is there an alternative proc or anything that I can use so I can still use the macro?

Thanks in advance!

Re: Alternative to proc iml?

RW9 — Mon, 14 Apr 2014 13:23:23 GMT

Best to post your code and example of what data you have and what you would like it to look like.

Re: Alternative to proc iml?

Rick_SAS — Mon, 14 Apr 2014 13:29:28 GMT

Short answer is "no, there isn't an easy replacement." Usually people incorporate SAS/IML because they want to perform matrix computations that are not readily available in Base SAS or SAS/STAT.

However, as @RW9 says, a definitive answer is not possible without knowing what the SAS/IML code in the macro does.

Your Bio doesn't give your affiliation, but I'll mention that SAS/IML will becaome available to all academic students and faculty in May 2014. See this announcement: SAS/IML available to all students through SAS Analytics U - The DO Loop

Re: Alternative to proc iml?

Sternchen — Mon, 14 Apr 2014 13:32:25 GMT

I'll try 🙂

The part of the macro that uses proc iml and where it stops is:

proc iml;

call symput('anzahl_eintraege',left(char(nrow({&liste}))));

quit;

%let anzahl_eintraege= %eval(&anzahl_eintraege*1);

proc iml;

a = t(1:&anzahl_eintraege);

b = {&liste};

c = length(b);

create listfile var{a b c};

append;

close listfile;

quit;

I have a dataset with a variable "ops_ko" for which I have 100 columns, "ops_ko1", "ops_ko2" and so on to "ops_ko100".

If any single one of these columns within a row contains a certain value, I want a (newly added) dummy variable to say 1, else 0.

The macro is supposed to do that, except that my SAS does not seem to work with proc iml so it is not beeing executed 😞

Does that make it better...?

Re: Alternative to proc iml?

data_null__ — Mon, 14 Apr 2014 13:41:16 GMT

Show that with example, this is what I have and this is what I need, data. The have/need scenario.

Re: Alternative to proc iml?

RW9 — Mon, 14 Apr 2014 13:49:38 GMT

Well, if its simply checking each column for a value and assigning a result then try (it generates a dataset with 100 if statements - I have assumed your value is 99):

data _null_;

call execute('data want;

set have;

attrib result format=best.;

result=0');

do i=1 to 100;

call execute('if ops_ko'||put(i,1.)||'=99 then result=1;');

end;

call execute('run;');

run;

Re: Alternative to proc iml?

Rick_SAS — Mon, 14 Apr 2014 13:59:06 GMT

I assume that &liste is a comma separated list that looks like this

x1,abc2,pdq,x4

The first PROC IML simply creates a macro variable with the number of entries in the list. For my example, &anzahl_eintraege will have the value 4.

I don't think that the %LET statement does anything useful.

The second PROC IML statement creates a data set named LISTFILE that contains three variables, A, B, and C.

A is just a counter. B is the name of the variables. C is the number of characters in the variable name. For my example, you'd get

A B C

1 x1 2

2 abc2 4

3 pdq 3

4 x1 2

You can replace this with a DATA step. I'll let someone else give you that code.

Re: Alternative to proc iml?

Sternchen — Mon, 14 Apr 2014 14:00:21 GMT

That is an example of my data (as I posted before, there is a total of 100 ops_ko-columns)

What I need is to add another variable that has the value 1 if in any of the ops_ko in a row has a certain value. For example if that value was 0107 then the newly added variable would be 1 for the first, second and fifth data row, if it was 9260 it would only be 1 for the second.

typ_alter	typ_geb	aufn_anl	aufn_grd	ops_ko1	ops_ko2	ops_ko3	ops_ko4	ops_ko5	ops_ko6	ops_ko7
18	2	N	0107			96030				80103
19	2	G	0107	9260		8800c0
10	2	E	0101	12750
14	2	E	0101	57583
17	2	N	0107	3222	1632
21	1	N	0501	87161