How to use Loops in Data manipulation in SAS - creating new columns

[imanojkumar1]

Hi all,

First of all I must thankful to everyone there who kept helping me to learn SAS fast. Almost a month ago, I was zero with SAS but now, at least I am feeling confident and have grown my love towards it. 🙂

Now, to my problem...

Suppose I have the following dataset:

data have;
  infile cards truncover expandtabs;
  input MC $ ET $ Date :Date9. Time :time8. Dia KMs;
  format Date date9. Time time8.;
cards;
US0001 Lath 12JAN15 7:00 890.8 2900334.00000
US000328 Lath 13JUL15 0:00 890.2 2232830.00000
US000328 Lath 02JAN16 0:00 885.9 2310299.00000
US000328 Lath 13JUN16 0:00 881.4 2450616.00000
US000329 Lath 12JAN15 0:00 885.7 2235626.00000
US000329 Lath 30MAY15 0:00 880.6 2373070.00000
US000329 Lath 14JAN16 0:00 872.6 2558987.00000
US000329 Lath 05AUG16 0:00 863.1 2740207.00000
US000331 Lath 13JUN15 0:00 860.7 2689105.00000
US000331 Lath 12JUN16 0:00 860.7 2730534.00000
US000334 Lath 02JUL15 0:00 859.6 3034370.00000
US000334 Lath 24JAN16 0:00 853.2 3230211.00000
US000334 Lath 27AUG16 0:00 845.9 3411050.00000
US000335 Lath 17MAR15 0:00 847.3 2183273.00000
US000335 Lath 15JAN16 0:00 847.3 2258321.00000
US000335 Lath 05AUG16 0:00 841.9 2438931.00000
US000336 Lath 16MAR15 0:00 853.6 2365989.00000
US000336 Lath 02MAY15 0:00 850.6 2405547.00000
US000336 Lath 28JAN16 0:00 846.3 2522017.00000
US000336 Lath 14MAR16 0:00 839.2 2554822.00000
;
run;

What I want to do, is that for each MC above:

1.  Calculate Dia difference in such a way that if there is only one row, keep the dia otherwise DiaDiff = Dia in current row - Dia in previous row (please Note: for practice, I used Case When Else option here, but you still can be better than me, I am sure)

2. If DiaDiff = Dia in that row, Dia_Flag = 0, otherwise =1

3. Difference between the Days in two rows. If first row, DaysDiff = 0, otherwise DaysDiff = Date in current row - Date in previous row

4. Similary for KMs Difference. If first row, KM_Diff = 0, otherwise KM_Diff = KMs in current row - KMs in previous row

And result would look like something below:

MC	ET	Date	Time	Dia	KMs	DiaDiff	Dia_Flag	DaysDiff	KM_Diff
US0001	Lath	12Jan2015	7:00:00	890,8	2900334	890,8	0	0	0
US000328	Lath	13Jul2015	0:00:00	890,2	2232830	890,2	0	0	0
US000328	Lath	02Jan2016	0:00:00	885,9	2310299	4,3	1	173	77469
US000328	Lath	13Jun2016	0:00:00	881,4	2450616	4,5	1	163	140317
US000329	Lath	12Jan2015	0:00:00	885,7	2235626	885,7	0	0	0
US000329	Lath	30May2015	0:00:00	880,6	2373070	5,1	1	138	137444
US000329	Lath	14Jan2016	0:00:00	872,6	2558987	8	1	229	185917
US000329	Lath	05Aug2016	0:00:00	863,1	2740207	9,5	1	204	181220
US000331	Lath	13Jun2015	0:00:00	860,7	2689105	860,7	0	0	0
US000331	Lath	12Jun2016	0:00:00	860,7	2730534	0	0	365	41429
US000334	Lath	02Jul2015	0:00:00	859,6	3034370	1,1	1	0	0
US000334	Lath	24Jan2016	0:00:00	853,2	3230211	6,4	1	206	195841
US000334	Lath	27Aug2016	0:00:00	845,9	3411050	7,3	1	216	180839
US000335	Lath	17Mar2015	0:00:00	847,3	2183273	847,3	0	0	0
US000335	Lath	15Jan2016	0:00:00	847,3	2258321	0	0	304	75048
US000335	Lath	05Aug2016	0:00:00	841,9	2438931	5,4	1	203	180610
US000336	Lath	16Mar2015	0:00:00	853,6	2365989	853,6	0	0	0
US000336	Lath	02May2015	0:00:00	850,6	2405547	3	1	47	39558
US000336	Lath	28Jan2016	0:00:00	846,3	2522017	4,3	1	271	116470
US000336	Lath	14Mar2016	0:00:00	839,2	2554822	7,1	1	46	32805

Currently I am learning loops and hence this case.

Thanks.

[RW9]

Hi,

Whilst I don't have time to do all that for you, some tips to get you started:

Use a datastep with a by group.

Use lag

E.g.

data want;
  set have;
  by mc;
  retain diadiff dia_flag daysdiff km_diff;
  if first.mc then do;
    diadiff=dia;
    dia=0;
    ...;
  end;
  else do;
    dia_flag=1;
    daysdiff=date-lag(date);
    ...;
  end;
run;

Note, I will repeat, its not a good idea to call variables type names e.g. date or time.

[Kurt_Bremser · Accepted solution]

Use the lag() function and first. condition to reset for every first observation in a group

data want;
set have;
by MC;
diadiff = dia - lag(dia);
daysdiff = date - lag(date);
km_diff = kms - lag(kms);
if first.MC
then do;
  diadiff = dia;
  daysdiff = 0;
  km_diff = 0;
end;
run;

If you want your difference values to always be positive, use the abs() function.

[imanojkumar1]

Thanks Kurt,

Only check in the above is, we need to use:
diadiff = lag(dia) - dia;

Otherwise the results will be negative numbers, if we use below:
diadiff = dia - lag(dia);