topic Re: Time to first event by group variable in SAS Programming

Time to first event by group variable

marinalev — Tue, 04 Aug 2020 20:29:42 GMT

I have a data set with the variables: participant ID (UUID), day, and event (infx_status). I would like to create a variable (DFI) that signifies when the first infection occurred for each participant. I'm having trouble especially when there is no infection present.

UUID	day	infx_status	DFI (want)
UU0333	0	0	2
UU0333	1	0	2
UU0333	2	1	2
UU0333	3	0	2
UU0333	4	1	2
UU0333	5	0	2
UU0334	1	0	5
UU0334	2	0	5
UU0334	3	0	5
UU0334	4	0	5
UU0334	5	1	5
UU0338	1	0	.
UU0338	2	0	.
UU0338	3	0	.
UU0338	4	0	.
UU0338	5	0	.

Don't know why I am having so much trouble with this but would appreciate some help. Thanks!

Re: Time to first event by group variable

ChrisNZ — Tue, 04 Aug 2020 20:57:29 GMT

Like this?

 select UUID, DAY, INFX_STATUS, min(DAY*ifn(INFX_STATUS, 1, .))
 from HAVE
 group by UUID  
 order by UUID, DAY;

UUID	day	INFX_STATUS
UU0333	0	0	2
UU0333	1	0	2
UU0333	2	1	2
UU0333	3	0	2
UU0333	4	1	2
UU0333	5	0	2
UU0334	1	0	5
UU0334	2	0	5
UU0334	3	0	5
UU0334	4	0	5
UU0334	5	1	5
UU0338	1	0	.
UU0338	2	0	.
UU0338	3	0	.
UU0338	4	0	.
UU0338	5	0	.

Re: Time to first event by group variable

marinalev — Tue, 04 Aug 2020 21:50:09 GMT

That's not working. The "min" statement is not changing color and I don't think that would create a new variable, would it? Sorry, I'm new to this.

Re: Time to first event by group variable

ChrisNZ — Tue, 04 Aug 2020 21:55:33 GMT

1. Run this in proc sql.

2. The color parser in EG is unreliable. Like the rest of EG. 😉

3. A new column is created, give it a name if you want: min(..) as VARNAMEOFYOURCHOICE

Re: Time to first event by group variable

mkeintz — Wed, 05 Aug 2020 19:11:04 GMT

The data step approach, if you dataset is sorted by UUID/DAY would be:

data want;
  set have (where=(infx_status=1) in=in1)
      have (in=in2);
  by id;
  retain dfi;
  if first.id then dfi=.;
  if in1=1 and dfi=. then dfi=day;
  if in2;
run;

Re: Time to first event by group variable

ChrisNZ — Tue, 04 Aug 2020 22:14:29 GMT

Or like this:

data WANT;
  set HAVE (where=(INFX_STATUS=1))
      HAVE (in=IN2);
  by UUID;
  retain DFI;
  if first.UUID then DFI=ifn(INFX_STATUS, DAY, .);
  if IN2;
run;

Re: Time to first event by group variable

ChrisNZ — Tue, 04 Aug 2020 22:18:03 GMT

I assumed you knew SQL basics, apologies if you don't.

The full code needed:

proc sql; 
 create table WANT as
 select UUID, DAY, INFX_STATUS, min(DAY*ifn(INFX_STATUS, 1, .)) as DIF
 from HAVE
 group by UUID  
 order by UUID, DAY;

Re: Time to first event by group variable

marinalev — Wed, 05 Aug 2020 14:24:27 GMT

Thank you so much!! Life saver!