AVERAGE OF A VARIABLE NUMBER OF VALUES

[mahele]

Hi everyone.
My SAS version is 9.4 and I have a dataset with 128967 observations, for each obs i have a date variable (dd/mm/yyyy)  and 12 months of amounts (IMPO_1 -- IMPO_12). Here is a subset of my dataset:

Oss	DATA_DEC	IMPO_1	IMPO_2	IMPO_3	IMPO_4	IMPO_5	IMPO_6	IMPO_7	IMPO_8	IMPO_9	IMPO_10	IMPO_11	IMPO_12
1	15/01/2018	549,78	547,07	544,71	542,36	539,84	537,32	534,8	532,28	530,75	529,45	527,86	.
2	25/05/2009	3.707,41	3.701,80	3.695,14	3.689,64	3.719,15	3.748,20	3.742,56	3.737,02	3.725,50	3.742,90	3.731,60	3.726,50
3	10/07/2009	307,2	306,4	301,5	310,5	312,23	309,45	305,12	301,23	307,25	289,56	286,25	215,9
4	15/04/2014	53,35	53,06	51,07	51,35	57,35	55,01	51,26	.	.	.	.	.
5	05/03/2009	121,16	119,48	114	100,89	85,82	76,43	82,31	76,71	76,25	74,69	75,14	73,12

For each observation I need to calculate the average of a variable number of 'IMPO_', this number of 'IMPO_' depends on month(DATA_DEC).

For example:

- for the first obs the         MONTH(DATA_DEC)=1 so AVERAGE=MEAN(IMPO_1);

- for the second obs the   MONTH(DATA_DEC)=5 so AVERAGE=MEAN(OF IMPO_1 -- IMPO_5);

- for the third obs the        MONTH(DATA_DEC)=7 so AVERAGE=MEAN(OF IMPO_1 -- IMPO_7);

.... and so on for each of 128967 observations.

My wanted dataset needs to be like:

Oss	DATA_DEC	IMPO_1	IMPO_2	IMPO_3	IMPO_4	IMPO_5	IMPO_6	IMPO_7	IMPO_8	IMPO_9	IMPO_10	IMPO_11	IMPO_12	AVERAGE
1	15/01/2018	549,78	547,07	544,71	542,36	539,84	537,32	534,8	532,28	530,75	529,45	527,86	.	549,78
2	25/05/2009	3.707,41	3.701,80	3.695,14	3.689,64	3.719,15	3.748,20	3.742,56	3.737,02	3.725,50	3.742,90	3.731,60	3.726,50	3.702,63
3	10/07/2009	307,2	306,4	301,5	310,5	312,23	309,45	305,12	301,23	307,25	289,56	286,25	215,9	307,49
4	15/04/2014	53,35	53,06	51,07	51,35	57,35	55,01	51,26	.	.	.	.	.	52,21
5	05/03/2009	121,16	119,48	114	100,89	85,82	76,43	82,31	76,71	76,25	74,69	75,14	73,12	118,21

How can i do this? I've tried creating a macro variable for each row, which holds the value of the month such as &month1, &month2...   then referencing it in to the MEAN(OF IMPO_1 -- IMPO_&&&month&n) but for 128967 rows this is madness.

Could someone help me please!!!
Thank you!

[novinosrin · Accepted solution]

data want;
 set have;
array t impo_1-impo_12;
 do i=1 to dim(t) until(i=month(data_dec));
 sum=sum(sum,t(i));
end;
avg=sum/i;
drop i;
run;

/*Or just*/

data want;
 set have;
array t impo_1-impo_12;
 do i=1 to month(data_dec);
 sum=sum(sum,t(i));
end;
avg=sum/i;
drop i;
run;

[mahele]

thank you so much!
the idea of using arrays is awsome! i introduced some  code changes  to manage  missing values.

data want;
	set have;
	array t impo_1-impo_12;

	do i=1 to month(data_dec);
		sum=sum(sum, t(i));

		if t(i) <>. then
			den=i;
		avg=round(sum/den,0.01);
	end;
	drop i sum den;
run;

and it works perfectly! thank you so much!

[Astounding]

Good idea to use arrays, but ....

The two variations look like they should generate different answers. The denominator in the bottom solution is wrong.

Also, it may be important to consider whether the data might contain missing values for some of the months. Again, the denominator would be incorrect if that happens.

[ballardw]

Assuming that your data_dec is a SAS date value:

data want;
 set have;
 select (month(data_dec));
   when (1) average= mean(impo_1);
   when (2) average= mean(of impo_1 -impo_2);
   when (3) average= mean(of impo_1 -impo_3);
   when (4) average= mean(of impo_1 -impo_4);
   when (5) average= mean(of impo_1 -impo_5);
   when (6) average= mean(of impo_1 -impo_6);
   when (7) average= mean(of impo_1 -impo_7);
   when (8) average= mean(of impo_1 -impo_8);
   when (9) average= mean(of impo_1 -impo_9);
   when (10) average= mean(of impo_1 -impo_10);
   when (11) average= mean(of impo_1 -impo_11);
   when (12) average= mean(of impo_1 -impo_12);
   otherwise;
 end;
run;

The Select statement in this case evaluates an expression month(data_dec) and branches to the WHEN value that matches.

I would be careful of using the two dash list indicator in case anything changes the order of your data. You could end up with other values in the calculations. When you know that have sequentially named values it is better to use the single dash for the list.

[mahele]

thank you!! it works great! thank you also for the sugestions regarding the two dash list indicator.