Transforming a dataset from having categorical data values (the date column) into a dataset with the same values as metadata (column names one per year) can be problematic for additional downstream processing.
The usual consumption is to create a report in the cross dimensional form and leave things as that.
For example:
data have (label="5 years of monthly values");
do date = '01jan2015'd to '31dec2019'd;
date = intnx ('month', date, 0, 'E');
percent = (year(date) + month(date) / 100) / 1e4;
output;
end;
format date yymmd8.;
format percent 8.6;
run;
ods listing;
options pagesize=21 nocenter nodate nonumber; title;
proc report data=have panels=3;
columns date percent;
define percent / display;
run;
------------------------- output -------------------------
date percent date percent date percent
2015-01 0.201501 2016-09 0.201609 2018-05 0.201805
2015-02 0.201502 2016-10 0.201610 2018-06 0.201806
2015-03 0.201503 2016-11 0.201611 2018-07 0.201807
2015-04 0.201504 2016-12 0.201612 2018-08 0.201808
2015-05 0.201505 2017-01 0.201701 2018-09 0.201809
2015-06 0.201506 2017-02 0.201702 2018-10 0.201810
2015-07 0.201507 2017-03 0.201703 2018-11 0.201811
2015-08 0.201508 2017-04 0.201704 2018-12 0.201812
2015-09 0.201509 2017-05 0.201705 2019-01 0.201901
2015-10 0.201510 2017-06 0.201706 2019-02 0.201902
2015-11 0.201511 2017-07 0.201707 2019-03 0.201903
2015-12 0.201512 2017-08 0.201708 2019-04 0.201904
2016-01 0.201601 2017-09 0.201709 2019-05 0.201905
2016-02 0.201602 2017-10 0.201710 2019-06 0.201906
2016-03 0.201603 2017-11 0.201711 2019-07 0.201907
2016-04 0.201604 2017-12 0.201712 2019-08 0.201908
2016-05 0.201605 2018-01 0.201801 2019-09 0.201909
2016-06 0.201606 2018-02 0.201802 2019-10 0.201910
2016-07 0.201607 2018-03 0.201803 2019-11 0.201911
2016-08 0.201608 2018-04 0.201804 2019-12 0.201912
Create a report
The TABULATE procedure needs a separate variable for each dimensional use (specified with CLASS statement). Since the same date value is contributing to the across and down dimensions, the date value needs to be repeated. This can be done with a data set view and thus would not required copying the original data into a new data set.
The formatted values of the class variables are used in the tabulation rendering.
data have2 / view=have2;
set have;
date2 = date;
run;
proc tabulate data=have2;
class date date2;
format date monname3.;
format date2 year.;
var percent;
table date='', date2='' * percent='' * sum='' * f=8.6;
run;
Data transformation
If you still want an actual data set with the year columns, the TRANSPOSE procedure would also require incoming data the date value in two forms -- the first as month (for BY processing that becomes the row 'id' and the second as year that become the column name via ID statement.
proc sql;
create view have_sorted_by_month_year as
select
put(date,monname3.) as month_name,
year(date) as year,
percent
from have
order by month(date), year(date);
quit;
proc transpose data=have_sorted_by_month_year out=want(drop=_name_);
by month_name notsorted;
var percent;
id year;
run;
proc print data=want;
run;
----------------------------------
month_
Obs name _2015 _2016 _2017 _2018 _2019
1 Jan 0.201501 0.201601 0.201701 0.201801 0.201901
2 Feb 0.201502 0.201602 0.201702 0.201802 0.201902
3 Mar 0.201503 0.201603 0.201703 0.201803 0.201903
4 Apr 0.201504 0.201604 0.201704 0.201804 0.201904
5 May 0.201505 0.201605 0.201705 0.201805 0.201905
6 Jun 0.201506 0.201606 0.201706 0.201806 0.201906
7 Jul 0.201507 0.201607 0.201707 0.201807 0.201907
8 Aug 0.201508 0.201608 0.201708 0.201808 0.201908
9 Sep 0.201509 0.201609 0.201709 0.201809 0.201909
10 Oct 0.201510 0.201610 0.201710 0.201810 0.201910
11 Nov 0.201511 0.201611 0.201711 0.201811 0.201911
12 Dec 0.201512 0.201612 0.201712 0.201812 0.201912
