cumulative sum by group on different conditions

[CHL0320]

Hello

I need suggestions on the following request.

New varname	Calculation
kcal_day	= sum of kcal
discsav_tot	= sum of disc_sav
discsav_kcal	= sum of kcal for all observations where disc_sav=1
discswt_tot	= sum of disc_swt
Discswt_kcal	= sum of kcal for all observations where disc_swt=1

The original data set is 

patid	date	kcal	disc_sav	disc_swt
1	13-Dec-14	124.8	0	.
1	13-Dec-14	278.64	0	1
1	13-Dec-14	0	.	.
1	13-Dec-14	58.5	0	.
1	13-Dec-14	1413.255	1	0
1	13-Dec-14	293.76	0	.
1	13-Dec-14	142.59	1	0
1	13-Dec-14	124.8	0	.
1	13-Dec-14	0	.	.
2	7-Jan-15	144	1	1
2	7-Jan-15	62.424	1	.
2	30-Mar-15	80.6	0	0
2	30-Mar-15	122	0	.
2	30-Mar-15	298.89	1	.
2	30-Mar-15	0	.	.
2	31-Mar-15	69.16	1	0
2	31-Mar-15	33.756624	0	.
2	31-Mar-15	51.072	1	1
2	31-Mar-15	62.62	1	.
2	31-Mar-15	0	.	.
2	31-Mar-15	14.025	1	.
2	30-Apr-15	10.35	1	.
2	30-Apr-15	85.92	1	.
2	30-Apr-15	26.555344	1	0
2	30-Apr-15	105.28	1	.

The final output wants only one raw per patid and date.  Thank you for your suggestion.

[Reeza · Accepted solution]

PROC MEANS + CLASS statement or PROC SQL. In general, PROC MEANS is more dynamic, while PROC SQL ends up being hard coded for your situation.