topic Re: create dataset with 3 periods in SAS Programming

create dataset with 3 periods

cuan — Tue, 01 Dec 2020 00:18:24 GMT

Dear SAS community,

I have a problem, it will be great if you can give me the solution.

I need to setup 3 periods:

dataset I have:

id date treat

1 10aug2019 1

1 16aug2019 1

1 20aug2019 1

1 30aug2019 1

1 30aug2019 2

1 02sep2019 2

1 06sep2019 2

1 09sep2019 2

1 15 sep2019 3

I need three periods:

dataset I want:

id period startdt stopdt

1 1 10 aug2019 29aug2019

1 2 30aug2019 09sep2019

1 3 10sep2019 15sep2019

Could you help me wo create these a dataset with 3 startdt and 3 stopdt?

Many thanks.

Cuan.

Re: create dataset with 3 periods

mkeintz — Tue, 01 Dec 2020 01:12:12 GMT

I presume period corresponds to treatment, except:

If two periods have an overlapping day, you want the overlapping date to belong exclusively to the later period, correct?
And if there is a gap between periods, you want the gap to belong to the later period. Is that also correct?

If so, then:

data have;
  input id date :date9. treat;
  format date date9.;
datalines;
1 10aug2019 1
1 16aug2019 1
1 20aug2019 1
1 30aug2019 1
1 30aug2019 2
1 02sep2019 2
1 06sep2019 2
1 09sep2019 2
1 15sep2019 3
run;

data want (keep=id treat start_dt end_dt);
  set have (keep=id treat) end=end_of_have;
  by id treat;

  merge have have (firstobs=2 keep=date rename=(date=_nxt_date));

  retain start_dt end_dt;
  format start_dt end_dt date9.;

  if first.treat then start_dt=ifn(first.id,date,sum(end_dt,1));
  if end_of_have then end_dt=date;
  else end_dt=min(date,_nxt_date-1);

  if last.treat;

run;

~~Note this only works if the overlap is never more than 1 day.~~ Reviewed the program - it DOES work if the overlap is more than 1 day.

Re: create dataset with 3 periods

cuan — Tue, 01 Dec 2020 01:15:22 GMT

Hi , many thanks for this,

but look like your code doesn't work with this dataset:

01AUG2017 1
16AUG2017 1
30AUG2017 1
30AUG2017 2
31AUG2017 2
01SEP2017 2
02SEP2017 2
03SEP2017 2
14SEP2017 3

It is getting the 3rd period: statdt=04sep2017 (correct, but stopdate=28nov2016 (incorrect),

it should be 14sep2017.

Not surre why you are getting 28nov2016.

Many tanks.

Re: create dataset with 3 periods

cuan — Tue, 01 Dec 2020 01:26:04 GMT

ohh I see,

your code is working fine, what about tif we put two id?

id date treat

1 01AUG2017 1
1 16AUG2017 1
1 30AUG2017 1
1 30AUG2017 2
1 31AUG2017 2
1 01SEP2017 2
1 02SEP2017 2
1 03SEP2017 2
1 14SEP2017 3
2 29NOV2016 1
2 13DEC2016 1
2 27DEC2016 1

please could you your code with now 2 id? because if you run your actual code, with these two id is what stopdate (3 period) come from = 29NOV2016, and should be 14sep2017.

Many thanks.

Re: create dataset with 3 periods

cuan — Tue, 01 Dec 2020 01:51:56 GMT

And if there is a gap between periods, you want the gap to belong to the later period. Is that also correct?

No, I ma afraid, this is not correct, I would like to consider gaps between periods, for example , like this example for one ID:

treatment 3: should start the 7jun, and your code start the 29May.

08MAR2018 1
23MAR2018 1
26APR2018 1
10MAY2018 1
24MAY2018 1
24MAY2018 2
26MAY2018 2
27MAY2018 2
28MAY2018 2
29MAY2018 2
07JUN2018 3
21JUN2018 3
21JUN2018 4
22JUN2018 4
23JUN2018 4
24JUN2018 4
25JUN2018 4
05JUL2018 5

Re: create dataset with 3 periods

mkeintz — Tue, 01 Dec 2020 02:00:32 GMT

@cuan wrote:

And if there is a gap between periods, you want the gap to belong to the later period. Is that also correct?

No, I ma afraid, this is not correct, I would like to consider gaps between periods, for example , like this example for one ID:

The last two observations of your initial problem statement had:

1 09sep2019 2

1 15 sep2019 3

which has 5 unaccounted-for days (10sep-14sep) between treatments 2 and 3 - i.e. a gap.

Yet you stated that you expected these dates for treatments 2 and 3:

id period startdt stopdt

1 2 30aug2019 09sep2019

1 3 10sep2019 15sep2019

This is clearly an example of assigning the gap to the later treatment. Now you've introduced some mission creep. Please take a long look at what you want and confirm before we attempt any more suggestions. Help us help you - state the task clearly and precisely.

Re: create dataset with 3 periods

cuan — Tue, 01 Dec 2020 02:11:31 GMT

my apologize for the confusion.

Will be great if you can solve my problem:

08MAR2018 1
23MAR2018 1
26APR2018 1
10MAY2018 1
24MAY2018 1
24MAY2018 2
26MAY2018 2
27MAY2018 2
28MAY2018 2
29MAY2018 2
07JUN2018 3
21JUN2018 3
21JUN2018 4

start date stop date

08mar2018 23may2018

24may2018 29may2018

07jun2018 20jun2018

21jun2018 21jun2018

Cheers,

Cuan.