Proc IML - evaluating multiple conditions

kc — Sun, 31 Jul 2022 04:45:59 GMT

I am new to using proc iml and have the following expression that doesn't seem to be working which I believe should be equivalent to 'if-then-else' statements in the data step. The variable 'teventBX' should be assigned a value (either period1/2/3/4/5) based on mutually exclusive conditional expressions.

teventBX = period1#(period1 < p1duration) + period2#((period1 > p1duration) & (1 <= period2 < 2))
+ period3#((period1 > p1duration) & (period2 > (p1duration + p2duration)) & (2 <= period3 < 3))
+ period4#((period1 > p1duration) & (period2 > (p1duration + p2duration)) & (period3 > (p1duration + p2duration + p3duration)) & (3 <= period4 < 4))
+ period5#((period1 > p1duration) & (period2 > (p1duration + p2duration)) & (period3 > (p1duration + p2duration + p3duration)) & (period4 > (p1duration + p2duration + p3duration + p4duration)) & (4 <= period5 < 5))
+ period5#((period1 > p1duration) & (period2 > (p1duration + p2duration)) & (period3 > (p1duration + p2duration + p3duration)) & (period4 > (p1duration + p2duration + p3duration + p4duration)) & (period5 > 5));

A few records in the dataset to help understand the issue. The last column has the correct values that should have been the result of the above code. But, except for the first 2 expressions for period1 and period2, the statement above is just adding up the period values for periods 3-5. For example, in Obs3 below, teventBX = period2 + period3 + period4 + period5. And in Obs4, teventBX = period2 + period3 + period4.

Obs	period1	p1duration	period2	p2duration	period3	p3duration	period4	p4duration	period5	p5duration	teventBX	Correct teventBX
1	1.7561	1	1.2074	1	5.9824	1	10.7529	1	4.6408	1	1.207	1.2074
2	0.6808	1	9.4134	1	7.1525	1	3.6021	1	4.7635	1	0.681	0.6808
3	1.7161	1	33.2933	1	42.1225	1	10.5065	1	4.1977	1	90.1200	4.1977
4	7.5841	1	16.5281	1	11.4061	1	3.8003	1	15.0174	1	31.7345	3.8003
5	10.4676	1	26.2526	1	2.0952	1	13.1565	1	4.8462	1	28.3478	2.0952
6	1.4104	1	4.0418	1	6.8128	1	9.4646	1	4.8171	1	25.1362	4.8171
7	4.2862	1	8.7941	1	2.5706	1	5.3440	1	7.3122	1	11.3647	2.5706
8	1.4407	1	2.1727	1	13.6600	1	37.2924	1	4.4550	1	57.5800	4.4550
9	7.8799	1	3.1210	1	4.2688	1	3.9127	1	11.4282	1	11.3025	3.9127

Any help is appreciated!

Re: Proc IML - evaluating multiple conditions

sbxkoenk — Sun, 31 Jul 2022 09:39:34 GMT

Hello,

There's a separate board for
SAS/IML Software and Matrix Computations
Statistical programming, matrix languages, and more
(Under the "Analytics" header).

I have moved your question to that place.

BR,

Koen

Re: Proc IML - evaluating multiple conditions

Rick_SAS — Sun, 31 Jul 2022 11:03:02 GMT

The expressions

(1 <= period2 <= 2)

and

(2 <= period3 < 3)

are not valid IML expressions. They should be replaced by

(1 <= period2) & (period2 <= 2)

and

(2 <= period3) & (period3 < 3)

To learn more about the BETWEEN operator in the DATA step and why it is not supported in MIL (or R or MATLAB....) see

https://blogs.sas.com/content/iml/2012/02/20/the-data-step-and-the-implied-and-operator.html

topic Re: Proc IML - evaluating multiple conditions in SAS/IML Software and Matrix Computations

Proc IML - evaluating multiple conditions

Re: Proc IML - evaluating multiple conditions

Re: Proc IML - evaluating multiple conditions