SAS dynamic arrays with changing start and stop values - arrays

Question

I am working on a longitudinal dataset where each row is a subject and each column is an event. There are no limits to the number of events that a subject can have, but the events are coded in a few ways. For the sake of this example let's say one of the coded ways is binary (good, bad).
I am trying to find
1) all of the event strings consisting of 3 or more events (with no count limit) that are within 24 hours of one another (on the same subject) from start to finish. There also may be multiple successes of this criteria within the same subject.
2) for each of the successes (string of 3 or more events within 24 hours) I need to count the number of good events.
I've included code that generates similar data to mine. For now I'm simplifying to 26 observations but I have up to 42 for a single subject.
data examp;
informat subject 4. epdt1 epdt2 epdt3 epdt4 epdt5 epdt6 epdt7 epdt8 epdt9 epdt10 epdt11 epdt12 epdt13 epdt14 epdt15 epdt16 epdt17 epdt18 epdt19 epdt20 epdt21 epdt22 epdt23 epdt24 epdt25 epdt26 datetime20.
good1 good2 good3 good4 good5 good6 good7 good8 good9 good10 good11 good12 good13 good14 good15 good16 good17 good18 good19 good20 good21 good22 good23 good24 good25 good26 1.;
input subject epdt1 epdt2 epdt3 epdt4 epdt5 epdt6 epdt7 epdt8 epdt9 epdt10 epdt11 epdt12 epdt13 epdt14 epdt15 epdt16 epdt17 epdt18 epdt19 epdt20 epdt21 epdt22 epdt23 epdt24 epdt25 epdt26
good1 good2 good3 good4 good5 good6 good7 good8 good9 good10 good11 good12 good13 good14 good15 good16 good17 good18 good19 good20 good21 good22 good23 good24 good25 good26;
format subject: 4. epdt: datetime20. good: 1.;
datalines;
3098 . . 25JUL1998:01:46:27 25JUL1998:02:16:05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3021 13JAN1999:17:31:37 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1982 01FEB1998:02:29:01 12APR1999:19:49:00 03JUN2018:21:00:00 13AUG1999:13:39:00 . . . . . . . . . . . . . . . . . . . . . . 1 . . . . . . . . . . . . . . . . . . . . . . . . .
1093 11APR2015:16:10:57 30AUG2015:00:52:28 14SEP2015:08:24:25 09MAY1999:00:28:37 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4089 29JUN1998:05:18:34 23JUL1998:18:31:11 07FEB1999:05:25:45 07FEB1999:05:29:26 07FEB1999:05:32:04 07FEB1999:05:34:05 14FEB1999:18:00:13 14FEB1999:18:01:02 14FEB1999:18:03:24 14FEB1999:18:05:55 14FEB1999:18:16:45 14FEB1999:18:19:04 14FEB1999:18:31:57 14FEB1999:18:35:22 28JUL1998:18:32:02 31DEC1998:00:22:33 . . . . . . . . 1 1 1 1 1 1 1 1 1 1 1 . 1 . 1 . . . . . . . . . .
3055 18FEB1998:11:34:00 14JUL1998:01:20:34 13OCT1998:10:49:08 30OCT1998:18:14:58 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1239 07MAR1998:06:02:18 01JUN1998:08:18:20 23JUN1998:07:52:11 04JUL1998:08:47:04 29JUL1998:23:16:41 29JUL1998:23:30:03 29JUL1998:23:42:56 30JUL1998:00:08:03 30JUL1998:00:12:30 30JUL1998:00:14:58 30JUL1998:00:36:00 30JUL1998:00:38:33 30JUL1998:00:57:56 30JUL1998:01:01:03 30JUL1998:01:06:10 30JUL1998:01:16:50 30JUL1998:01:24:19 30JUL1998:01:32:30 30JUL1998:01:42:55 30JUL1998:01:50:24 30JUL1998:02:08:46 30JUL1998:02:20:18 30JUL1998:02:22:08 30JUL1998:02:28:52 30JUL1998:02:31:29 30JUL1998:02:51:29 . . 1 . 1 1 1 1 1 1 1 . 1 1 1 1 1 1 1 1 1 1 1 1 . 1
9834 10JUL1999:20:22:24 14JUL1999:00:52:02 14JUL1999:17:02:38 14JUL1999:17:30:06 21FEB2000:12:41:34 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
run;
proc sort data=examp; by subject;
data epwide_dt1;
format apppair $7000.;
set examp;
by subject;
%macro loops;
array eptm (*)epdt1-epdt26; array apptm (*) good1-good26;
*********using the starting value for identifying pairs;
*******trimmed then for the sake of making the macro work;
%do start=1 %to 26;
%do stop=3 %to 26;
%if &start.<&stop. %then %do ;
/***********to figure out if the difference between the pairs of times are 24 hours;*/
tbtw=eptm[&stop.]-eptm[&start.];
/* *********number of points between them;*/
diff=(&stop.)- (&start.);
*******calculate the summaries between all episodes from start to stop;
array appr&start.&stop. (*) ap&start.-ap&stop.;
array stmct&start.&stop.(*) st&start.-st&stop.;
%do i=&start. %to &stop.;
******calculate the number of appropriate episodes;
if apptm[&i] ne . then appr&start.&stop.[&i]=apptm[&i];
else appr&start.&stop.[&i]=0;
totapp=sum(of appr&start.&stop.(*));
if totapp=. then totapp=0;
****after you calculate the total value dump the array before the next itteration;
/*call missing(of appr&start.&stop.{*});*/
if (eptm[&start.] ne . and eptm[&stop.] ne . and diff>=2 and .<tbtw<86400 and totapp>1 ) then do;
appPair=catx(" ",apppair,"(",strip(put(&start., 3.)),"-",strip(put(&stop.,3.)),":", strip(put(totapp,3.)),"Good)");
end;
%end;
%end;
%end;
%end;
%mend;
%loops ;
run;
The error message below is what's resulting:
ERROR: Array subscript out of range at line 1 column 2.
apppair= subject=1093 epdt1=11APR2015:16:10:57 epdt2=30AUG2015:00:52:28 epdt3=14SEP2015:08:24:25
epdt4=09MAY1999:00:28:37 epdt5=. epdt6=. epdt7=. epdt8=. epdt9=. epdt10=. epdt11=. epdt12=. epdt13=. epdt14=. epdt15=.
epdt16=. epdt17=. epdt18=. epdt19=. epdt20=. epdt21=. epdt22=. epdt23=. epdt24=. epdt25=. epdt26=. good1=. good2=.
good3=. good4=. good5=. good6=. good7=. good8=. good9=. good10=. good11=. good12=. good13=. good14=. good15=. good16=.
good17=. good18=. good19=. good20=. good21=. good22=. good23=. good24=. good25=. good26=. FIRST.subject=1
LAST.subject=1 tbtw=1323117 diff=1 ap1=0 ap2=0 ap3=0 st1=. st2=. st3=. totapp=0 ap4=0 st4=. ap5=0 st5=. ap6=0 st6=.
ap7=0 st7=. ap8=0 st8=. ap9=0 st9=. ap10=0 st10=. ap11=0 st11=. ap12=0 st12=. ap13=0 st13=. ap14=0 st14=. ap15=0
st15=. ap16=0 st16=. ap17=0 st17=. ap18=0 st18=. ap19=0 st19=. ap20=0 st20=. ap21=0 st21=. ap22=0 st22=. ap23=0 st23=.
ap24=0 st24=. ap25=0 st25=. ap26=0 st26=. _ERROR_=1 _N_=1
NOTE: Missing values were generated as a result of performing an operation on missing values.
Each place is given by: (Number of times) at (Line):(Column).
1 at 35:20 1 at 57:20 1 at 83:20 1 at 113:20 1 at 147:20 1 at 185:20 1 at 227:20
1 at 273:20 1 at 323:20 1 at 377:20 1 at 435:20 1 at 497:20 1 at 563:20 1 at 633:20
1 at 707:20 1 at 785:20 1 at 867:20 1 at 953:20 1 at 1043:20 1 at 1137:20 1 at 1235:20
1 at 1337:20
NOTE: The SAS System stopped processing this step because of errors.
NOTE: There were 2 observations read from the data set WORK.EXAMP.
WARNING: The data set WORK.EPWIDE_DT1 may be incomplete. When this step was stopped there were 0 observations and
109 variables.
WARNING: Data set WORK.EPWIDE_DT1 was not replaced because this step was stopped.
NOTE: DATA statement used (Total process time):
real time 2.35 seconds
cpu time 2.13 seconds
Thanks in advance for any suggestions!

Answer 1

I am not sure I quite understand your full problem. But consider that if you want to sum the subset of values in an array from index START to index STOP you just use a DO loop.
For example to sum X10 to X20 you could use code like this:
array x (100) ;
start=10;
stop=20;
do i=start to stop;
total=sum(total,0,x(i));
end;
So you should be able to solve this without macro code. Which should make debugging much easier.

Answer 2

I finally got it working!!! I used #Tom 's suggestion to remove the need for creating a sub array of each of the pairs, as it was causing ALOT of problems. I also streamlined the output and asked it to output each pair that is "Good" so that I will be able to evaluate them easier. Previously it was creating appPair (my evaluation summary for each iteration within the start stop loop producing a bunch of extraneous output).
data epwide_dt1;
set examp;
by subject;
if first.subject then totapp=0;
%macro loops;
array eptm (*)epdt1-epdt26;
array apptm (*) good1-good26;
*********using the starting value for identifying pairs;
%do start=1 %to 24;
%do stop=3 %to 26;
%if &start.<&stop. %then %do ;
totapp=0;
/***********to figure out if the difference between the pairs of times are 24 hours;*/
tbtw=eptm[&stop.]-eptm[&start.];
/* *********number of points between them;*/
diff=(&stop.)- (&start.);
%do i=&start. %to &stop.;
******calculate the number of good events;
totapp=sum(totapp, 0,apptm[&i]);
***output the summary on the pair that can be evaluated in the next step;
if &i=&stop. and (eptm[&start.] ne . and eptm[&stop.] ne . and diff>=2 and 0<tbtw<86400 and totapp>1 ) then do;
appPair=catx(" ","(",strip(put(&start., 3.)),"-",strip(put(&stop.,3.)),":", strip(put(totapp,3.)),"Good)");
output;
end;
%end;
%end;
%end;
%end;
%mend;
%loops ;
run;