How do I keep the dataset's month in proc timeseries when converting to yearly frequency data?

[KALLEN]

I am converting daily data into yearly data for August 1999 - 2018. However, when I use proc timeseries, it starts the series at January 1, 1999 instead of at August 1, 1999.  How do I keep August and prevent proc timeseries from labeling the date as January?

Below is my code:

/* convert monthly data to yearly data */
proc timeseries data=aug_monthly_weather_90_100 out=aug_yearly_weather_90_100;
	id date interval=year accumulate=n;
	var Max_Temp Max_Temp_90 Max_Temp_100;
	format date date9.;
run;

Current, incorrect output:

Obs Date Max_Temp Max_Temp_90 Max_Temp_100

1	01JAN1999	31	25	0
2	01JAN2000	31	27	0
3	01JAN2001	31	25	1
4	01JAN2002	31	29	9
5	01JAN2003	31	27	5
6	01JAN2004	31	22	0
7	01JAN2005	31	19	0
8	01JAN2006	31	11	0
9	01JAN2007	31	27	3
10	01JAN2008	31	17	0
11	01JAN2009	31	29	3
12	01JAN2010	31	27	10
13	01JAN2011	31	30	15
14	01JAN2012	31	30	8
15	01JAN2013	31	26	2
16	01JAN2014	31	17	1
17	01JAN2015	31	31	10
18	01JAN2016	31	20	2
19	01JAN2017	31	25	0
20	01JAN2018	31	28	8

Desired output:

Obs Date Max_Temp Max_Temp_90 Max_Temp_100

1	01AUG1999	31	25	0
2	01AUG2000	31	27	0
3	01AUG2001	31	25	1
4	01AUG2002	31	29	9
5	01AUG2003	31	27	5
6	01AUG2004	31	22	0
7	01AUG2005	31	19	0
8	01AUG2006	31	11	0
9	01AUG2007	31	27	3
10	01AUG2008	31	17	0
11	01AUG2009	31	29	3
12	01AUG2010	31	27	10
13	01AUG2011	31	30	15
14	01AUG2012	31	30	8
15	01AUG2013	31	26	2
16	01AUG2014	31	17	1
17	01AUG2015	31	31	10
18	01AUG2016	31	20	2
19	01AUG2017	31	25	0
20	01AUG2018	31	28	8

I have attached my dataset.

[mkeintz · Accepted solution]

Does proc timeseries accept "interval=year.8"?   (meaning years that begin on the 8'th month).

[KALLEN]

Yes, a very quick and elegant solution. Thanks!