topic Re: Count column to a max value using other columns in New SAS User

Count column to a max value using other columns

1SasUser1 — Thu, 01 Aug 2019 09:09:45 GMT

Hi,

I am using SAS Enterprise Guide 5.1 version 9.3.

I am trying to create a new column by looking at other columns.

I have columns: amt_reached, amt_needed, Level and each are split by ID.

Currently if the amt_reached is not greater than or equal to amt_needed, then Level = 1, otherwise it is 0. I would like a new column to count Level up to 3 and then this is the max and all values after would be at 0 unless the following line the amt_reached is greater than or equals amt_needed. Then the count would restart.

Here is what I want:

ID	Level	amt_reached	amt_needed	What I want
123	0	231.71	482.73	1
123	1	251.02	235.1	2
123	1	235.1	476.33	3
123	1	241.23	478.95	0
123	1	237.72	470.33	0
123	0	465.77	465.77	0
123	1	.	215.89	1
123	1	215.89	460.7	2
123	1	244.81	486.22	3
123	1	241.41	477.02	0
123	1	235.61	469.03	0
123	1	233.42	475.1	0
234	1	139.35	278.7	1
234	1	139.35	278.7	2
234	1	139.35	278.7	3
234	1	139.35	278.7	0
234	1	139.35	278.7	0
234	1	139.35	278.7	0
234	1	139.35	278.7	0
234	1	139.35	278.7	0
234	1	139.35	278.7	0
234	1	139.35	278.7	0
234	1	139.35	278.7	0
234	1	139.35	278.7	0
156	1	.	18.99	1
156	1	18.99	39.07	2
156	1	20.08	40.26	3
156	1	20.18	40.46	0
156	1	20.28	40.63	0
156	1	20.35	40.8	0
156	1	20.45	40.97	0
156	0	61.64	41.12	0
156	1	.	18.02	1
156	0	37.09	37.09	0
156	1	.	17.68	1
156	1	17.68	36.4	2

SAS Code:

data WORK.COUNTWITHMAX;
  infile datalines dsd truncover;
  input ID:BEST12. Level:BEST12. amt_reached:BEST12. amt_needed:BEST12.;
  format ID BEST12. Level BEST12. amt_reached BEST12. amt_needed BEST12.;
datalines;
123 0 231.71 482.73
123 1 251.02 235.1
123 1 235.1 476.33
123 1 241.23 478.95
123 1 237.72 470.33
123 0 465.77 465.77
123 1 . 215.89
123 1 215.89 460.7
123 1 244.81 486.22
123 1 241.41 477.02
123 1 235.61 469.03
123 1 233.42 475.1
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
156 1 . 18.99
156 1 18.99 39.07
156 1 20.08 40.26
156 1 20.18 40.46
156 1 20.28 40.63
156 1 20.35 40.8
156 1 20.45 40.97
156 0 61.64 41.12
156 1 . 18.02
156 0 37.09 37.09
156 1 . 17.68
156 1 17.68 36.4
;;;;

Here is what I have tried so far:

retain MaxCount;
	if first.ID	
		then MaxCount = 0;
	if Level = 0
		then MaxCount = 0;
	else do;
		MaxCount = mod(MaxCount + 1,4);
if MaxCount = 0 and amt_reached >= amt_needed then MaxCount = 1;


end;

It does not currently work, but I can't quite finish it.

Any help would be greatly appreciated

Thanks!

Re: Count column to a max value using other columns

Kurt_Bremser — Thu, 01 Aug 2019 09:42:19 GMT

Per your description, you would restart with 1 in observation 5, as in observation 6 amt_reached equals amt_needed. Please clarify.

Re: Count column to a max value using other columns

1SasUser1 — Thu, 01 Aug 2019 09:52:38 GMT

Hi KurtBremser,

Thanks for your question.

As in observation 6 column amt_reached = amt_needed the new column remains at 0. It is observation 7 where the count can restart, not observation 5. If amt_reached in observation 6 had been less than amt_needed then the new column would remain at 0 until there was on occasion where amt_reached >= amt_needed.

The count always gets to 3 and remains at 0 for the next observations until amt_reached >= amt_needed.

Hope this clarifies.

Re: Count column to a max value using other columns

Kurt_Bremser — Thu, 01 Aug 2019 11:02:06 GMT

So you mean to start a new series when the previous observation satisfied the condition?

But then I do no understand the last 4 observations:

156	1	.	18.02	1
156	0	37.09	37.09	0
156	1	.	17.68	1
156	1	17.68	36.4	2

Should't those values be 1 - 2 - 1 - 2?

Re: Count column to a max value using other columns

Kurt_Bremser — Thu, 01 Aug 2019 11:08:17 GMT

I think I have found it. It creates your output:

data have;
input ID Level amt_reached amt_needed;
datalines;
123 0 231.71 482.73
123 1 251.02 235.1
123 1 235.1 476.33
123 1 241.23 478.95
123 1 237.72 470.33
123 0 465.77 465.77
123 1 . 215.89
123 1 215.89 460.7
123 1 244.81 486.22
123 1 241.41 477.02
123 1 235.61 469.03
123 1 233.42 475.1
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
234 1 139.35 278.7
156 1 . 18.99
156 1 18.99 39.07
156 1 20.08 40.26
156 1 20.18 40.46
156 1 20.28 40.63
156 1 20.35 40.8
156 1 20.45 40.97
156 0 61.64 41.12
156 1 . 18.02
156 0 37.09 37.09
156 1 . 17.68
156 1 17.68 36.4
;

data want;
set have;
by id notsorted;
_level = lag(level);
if first.id then new = 1;
else do;
  if level = 0 then new = 0;
  if new in (1,2) then new + 1;
  else if new = 3 then new = 0;
  else if _level = 0 then new = 1;
end;
run;

proc print data=want noobs;
run;

Result:

                  amt_      amt_
 ID    Level    reached    needed    _level    new

123      0       231.71    482.73       .       1 
123      1       251.02    235.10       0       2 
123      1       235.10    476.33       1       3 
123      1       241.23    478.95       1       0 
123      1       237.72    470.33       1       0 
123      0       465.77    465.77       1       0 
123      1          .      215.89       0       1 
123      1       215.89    460.70       1       2 
123      1       244.81    486.22       1       3 
123      1       241.41    477.02       1       0 
123      1       235.61    469.03       1       0 
123      1       233.42    475.10       1       0 
234      1       139.35    278.70       1       1 
234      1       139.35    278.70       1       2 
234      1       139.35    278.70       1       3 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
234      1       139.35    278.70       1       0 
156      1          .       18.99       1       1 
156      1        18.99     39.07       1       2 
156      1        20.08     40.26       1       3 
156      1        20.18     40.46       1       0 
156      1        20.28     40.63       1       0 
156      1        20.35     40.80       1       0 
156      1        20.45     40.97       1       0 
156      0        61.64     41.12       1       0 
156      1          .       18.02       0       1 
156      0        37.09     37.09       1       0 
156      1          .       17.68       0       1 
156      1        17.68     36.40       1       2

Re: Count column to a max value using other columns

1SasUser1 — Thu, 01 Aug 2019 12:15:05 GMT

Thanks that does work for these cases. I have found a case where it doesn't work and I think perhaps because there is no reference to the amt_reached and amt_needed columns.

In reference to your previous question for the last 4 observations, the reason why I want it to be 1,0,1,2 is because the 3rd from last observation is as follows:

ID	Level	amt_reached	amt_needed	What I want
156	0	37.09	37.09	0

Since the columns amt_reached and amt_needed fall are equal, then the new column will be at 0. If amt_reached < amt_needed, then it should be 1,2,3,0 for the last 4 observations. Whenever Level = 0 and amt_reached >= amt_needed then that row should = 0 and the count restarts at the next observation.

However if Level = 1 and amt_reached >= amt_needed, then the new column should equal 1 at that point. These two columns are the main decision points.

Hope that is a bit clearer.

Here is an example of when it doesn't work with your new code:

ID	_Level	new	Level	amt_reached	amt_needed	What I want
111	0	1	1	125.27	247.92	1
111	1	2	1	122.65	242.71	2
111	1	3	1	120.06	237.33	3
111	1	0	1	117.27	232.1	0
111	1	0	1	114.83	227.21	0
111	1	0	1	112.38	222.43	0
111	1	0	1	110.05	217.77	0
111	1	0	0	213.27	213.27	0
111	0	1	0	97.51	97.51	0
111	0	1	0	95.02	95.02	0
111	0	1	0	92.67	92.67	0
111	0	1	0	90.3	90.3	0

Thanks again!

Re: Count column to a max value using other columns

Kurt_Bremser — Thu, 01 Aug 2019 12:39:07 GMT

I think a small change in the code should do it:

data want;
set have;
by id notsorted;
_level = lag(level);
if first.id then new = 1;
else do;
  if level = 0 then new = 0;
  else if new in (1,2) then new + 1;
  else if new = 3 then new = 0;
  else if _level = 0 then new = 1;
end;
run;

Note the extra "else" in line 8 that makes it a whole if-then-else-if chain.

Re: Count column to a max value using other columns

1SasUser1 — Thu, 01 Aug 2019 13:01:50 GMT

That did the trick.

The other thing is with the code saying if first.ID then new = 1, this is causing some cases where a new ID begins to have 1 on the first observation, even if Level doesn't equal 1. I also can see cases where new = 2 but the previous observation isn't there is at 0

Please see below:

ID	_Level	new	Level	amt_reached	amt_needed	What I want
222	.	1	0	21	20.59	0
222	0	0	0	21	20.25	0
222	0	0	0	21	20.45	0
222	1	0	0	22.54	22.54	0
222	0	0	0	21	20.06	0
222	0	0	0	20	19.46	0
222	0	0	0	20	19.92	0
222	0	1	1	18.5	18.95	1
222	0	0	0	23	21.59	0
222	0	0	0	30	19.14	0
222	0	0	0	21	20.97	0
222	0	0	0	21	19.71	0
333	0	1	0	8.99	7.2	0
333	0	0	0	7.69	5.13	0
333	0	0	0	5.44	5	0
333	0	0	0	7.54	5	0
333	0	0	0	10.51	5.53	0
333	0	0	0	10	5.9	0
333	0	0	0	7.67	5.13	0
333	1	0	0	6.77	5.34	0
333	0	0	0	10	5	0
333	0	1	1	6	5	1
333	0	0	0	7.36	5	0
333	0	0	0	7	5	0
444	1	2	1	.	5	1
444	1	0	0	77.7	10	0
444	0	1	1	20	5	1
444	0	0	0	20	5	0
444	0	1	0	20	12.09	0
444	0	1	1	.	5	1
444	0	0	0	20	5.52	0
444	1	0	0	20	10.34	0
444	0	1	1	20	5	1
444	0	0	0	.	0	0
444	1	2	1	50	13.8	1

Re: Count column to a max value using other columns

1SasUser1 — Thu, 01 Aug 2019 13:52:54 GMT

Sorry, I realised I had ordered the data incorrectly for my last comment. The issue with new = 2 is fine. The issue is only with the starting ID = 1.

Apologies.