Simplify bloc of conditioons

[polpel]

Hello,

I'm new to Sas and was wondering if this code could be simplified:

I'm basically adding a column for the person's age range (5 year range), for other functional procedures i add afterwards, but this is the part that takes longest to process in my program (plus my god how ugly :D)

DATA tranches1;
	set DEMO;
	if Age < 18 then
		TraLibelle = "Moins de 18 ans";
	else if 18 < AGE <= 24 then
		TraLibelle = "De 18 à 24 ans";
	else if 24 < AGE <= 29 then
		TraLibelle = "De 25 à 29 ans";
	else if 29 < AGE <= 34 then
		TraLibelle = "De 30 à 34 ans";
	else if 34 < AGE <= 39 then
		TraLibelle = "De 35 à 39 ans";
	else if 39 < AGE <= 44 then
		TraLibelle = "De 40 à 44 ans";
	else if 44 < AGE <= 49 then
		TraLibelle = "De 45 à 49 ans";
	else if 49 < AGE <= 54 then
		TraLibelle = "De 50 à 54 ans";
	else if 54 < AGE <= 59 then
		TraLibelle = "De 55 à 59 ans";
	else if 59 < AGE <= 64 then
		TraLibelle = "De 60 à 64 ans";
	else
		TraLibelle = "65 ans et plus";
RUN;

[novinosrin]

Hi @polpel  IF THEN ELSE for your task is rather boring and tedious. Please take a look at USER DEFINED FORMATS aka PROC FORMAT. Do a simple search on the net, you will get gazillion examples. 

[polpel]

hi, thanks for the quick response.

I already have a format set so that i can later order by ascending age range, here's the code:

PROC FORMAT;
	value $tranche_age
		"Moins de 18 ans"=01
		"De 18 à 24 ans"=02
		"De 25 à 29 ans"=03
		"De 30 à 34 ans"=04
		"De 35 à 39 ans"=05
		"De 40 à 44 ans"=06
		"De 45 à 49 ans"=07
		"De 50 à 54 ans"=08
		"De 55 à 59 ans"=09
		"De 60 à 64 ans"=10
		"65 ans et plus"=11
	;
QUIT;

could i possibly use this to implement what i'm looking to do?

right now i was able to change my code to this:

DATA tranches1;
	set DEMO;
	select;
	when (Age < 18)					TraLibelle = "Moins de 18 ans";
	when (18 <= Age <= 24)			TraLibelle = "De 18 à 24 ans";
	when (24 <= Age <= 29)			TraLibelle = "De 25 à 29 ans";
	when (29 <= Age <= 34)			TraLibelle = "De 30 à 34 ans";
	when (34 <= Age <= 39)			TraLibelle = "De 35 à 39 ans";
	when (39 <= Age <=44)			TraLibelle = "De 40 à 44 ans";
	when (44 <= Age <=49)			TraLibelle = "De 45 à 49 ans";
	when (49 <= Age <=54)			TraLibelle = "De 50 à 54 ans";
	when (54 <= Age <=59)			TraLibelle = "De 55 à 59 ans";
	when (59 <= Age <=64)			TraLibelle = "De 60 à 64 ans";
	otherwise						TraLibelle = "65 ans et plus";
	end;
RUN;

[yabwon]

Hi @polpel ,

Try SELECT-WHEN statement:

1) nice blog post by DO-LOOP: https://blogs.sas.com/content/iml/2016/06/20/select-when-sas-data-step.html

2) documentation: https://documentation.sas.com/?docsetId=lestmtsref&docsetTarget=p09213s9jc2t99n1vx0omk2rh9ps.htm&doc...

All the best

Bart;

DATA tranches1;
	set DEMO;
  select ;
  	when ( Age <  18 ) TraLibelle = "Moins de 18 ans";
  	when ( AGE <= 24 ) TraLibelle = "De 18 à 24 ans";
  	when ( AGE <= 29 ) TraLibelle = "De 25 à 29 ans";
  	when ( AGE <= 34 ) TraLibelle = "De 30 à 34 ans";
  	when ( AGE <= 39 ) TraLibelle = "De 35 à 39 ans";
  	when ( AGE <= 44 ) TraLibelle = "De 40 à 44 ans";
  	when ( AGE <= 49 ) TraLibelle = "De 45 à 49 ans";
  	when ( AGE <= 54 ) TraLibelle = "De 50 à 54 ans";
  	when ( AGE <= 59 ) TraLibelle = "De 55 à 59 ans";
  	when ( AGE <= 64 ) TraLibelle = "De 60 à 64 ans";
  	otherwise TraLibelle = "65 ans et plus";
  end;
RUN;

[yabwon]

@polpel ,

One more approach would be double array and do-loop. But treat it more like "extra SAS tutorial" rather than real use case.

All the best

Bart

data Demo;
do Age = 7 to 77 by 3;
  output;
end;
run ;

DATA tranches2;
	set DEMO;

  array A[0:11] _temporary_ (0 18 24 29 34 39 44 49 54 59 64 999);
  array T[  11] $ 20 _temporary_ (
                    "Moins de 18 ans"
                    "De 18 à 24 ans"
                    "De 25 à 29 ans"
                    "De 30 à 34 ans"
                    "De 35 à 39 ans"
                    "De 40 à 44 ans"
                    "De 45 à 49 ans"
                    "De 50 à 54 ans"
                    "De 55 à 59 ans"
                    "De 60 à 64 ans"
                    "65 ans et plus"
                    );

  do _N_ = 1 to 11;
    if A[_N_-1] < Age <= A[_N_] then
      do;
        TraLibelle =  T[_N_] ;
        leave;
      end;
  end;

RUN;

[polpel]

hey!
this works, but only for ages 18 and above, age range from 0 to 18 doesnt appear in result table...

[yabwon]

strange, I have the following output in both cases (select+when and arrays):

Obs    Age      TraLibelle

  1      7    Moins de 18 ans
  2     10    Moins de 18 ans
  3     13    Moins de 18 ans
  4     16    Moins de 18 ans
  5     19    De 18 Ă  24 ans
  6     22    De 18 Ă  24 ans
  7     25    De 25 Ă  29 ans
  8     28    De 25 Ă  29 ans
  9     31    De 30 Ă  34 ans
 10     34    De 30 Ă  34 ans
 11     37    De 35 Ă  39 ans
 12     40    De 40 Ă  44 ans
 13     43    De 40 Ă  44 ans
 14     46    De 45 Ă  49 ans
 15     49    De 45 Ă  49 ans
 16     52    De 50 Ă  54 ans
 17     55    De 55 Ă  59 ans
 18     58    De 55 Ă  59 ans
 19     61    De 60 Ă  64 ans
 20     64    De 60 Ă  64 ans
 21     67    65 ans et plus
 22     70    65 ans et plus
 23     73    65 ans et plus
 24     76    65 ans et plus

Bart

[KachiM · Accepted solution]

Five year grouping is made for age 25 to 64. It can be used for fun as:

data have;
	do i = 1 to 20;
		age = int(99 * uniform(123));
		output;
	end;
drop i;
run;


data want;
array T[11] $ 20 _temporary_ (
                    "Moins de 18 ans"
                    "De 18 à 24 ans"
                    "De 25 à 29 ans"
                    "De 30 à 34 ans"
                    "De 35 à 39 ans"
                    "De 40 à 44 ans"
                    "De 45 à 49 ans"
                    "De 50 à 54 ans"
                    "De 55 à 59 ans"
                    "De 60 à 64 ans"
                    "65 ans et plus"
                    );
	set have;
	if age < 18 then i = 1;
	else if age < 25 then i = 2;
	else if age > 64 then i = 11;
	else i = 2 + ceil((age - 24) / 5);
	TraLibelle = T[i];
drop i;
run;

[polpel]

hi, thanks for response,

only i don't understand what you're doing in the first step...

[KachiM]

data want;
array T[11] $ 20 _temporary_ (
                    "Moins de 18 ans"
                    "De 18 à 24 ans"
                    "De 25 à 29 ans"
                    "De 30 à 34 ans"
                    "De 35 à 39 ans"
                    "De 40 à 44 ans"
                    "De 45 à 49 ans"
                    "De 50 à 54 ans"
                    "De 55 à 59 ans"
                    "De 60 à 64 ans"
                    "65 ans et plus"
                    );
	set have;
	if age < 18 then i = 1;
	else if age < 25 then i = 2;
	else if age > 64 then i = 11;

I am not sure what you mean by "----- i don't understand what you're doing in the first step...".

However, I will explain the logic used.

You require grouping of ages - a total of 11 groups. Each group is labelled and stored in an array T[ ]. Ages 25 to 64 need be grouped into 5-year widths which means 8 intervals(i=3 to i=10). Ages 18 to 24 has a width of 6 years. Rest of the years are falling in open intervals.

First interval i = 1 takes ages below 18 years.

Ages 18 to 24, 6-year width, will be placed in 2nd interval (i = 2).

Ages 65+ will be placed in i = 11.

Ages 25 to 64 takes intervals as said above. The corresponding i (i=3, ...,10) is computed using:

else i = 2 + ceil((age - 24) / 5);

Hope this clarifies your query.

Kind regards,

DataSP

[polpel]

Hello again and thanks for the quick answers!

I was able to use your method but I get an error at the end...

Here's the complete code

DATA calc;
	array T[11] $ 20 _temporary_ (
                    "Moins de 18 ans"
                    "De 18 à 24 ans"
                    "De 25 à 29 ans"
                    "De 30 à 34 ans"
                    "De 35 à 39 ans"
                    "De 40 à 44 ans"
                    "De 45 à 49 ans"
                    "De 50 à 54 ans"
                    "De 55 à 59 ans"
                    "De 60 à 64 ans"
                    "65 ans et plus"
                    );
	Lotdeselection = &NUM_CTR.;
	set DEMO (where=(type_ben = "Adhérent"));
	if age < 18 then i = 1;
	else if age < 25 then i = 2;
	else if age > 64 then i = 11;
	else i = 2 + ceil((age - 24) / 5);
	TraLibelle = T[i];
drop i;
RUN;

proc tabulate data=calc /*format=Z3.1*/out=tbdN /*style=[1]*/; 
	class TraLibelle sexe LotdeSelection; 
	var prorata; 
	table TraLibelle*LotdeSelection,sexe*(prorata)*(sum /*colpctsum*F=7.2*/); 
run; 
ODS SELECT ALL ;  
proc sort data=tbdN (drop=_page_ _TYPE_ _TABLE_); 
	by  TraLibelle; 
run; 
proc transpose data=tbdN out=T_; 
	by TraLibelle LotdeSelection; 
	var prorata_Sum;
	id SEXE; 
run; 
proc sql; 
	create table tbd5 as 
		select LotdeSelection, 
			TraLibelle,/*put(TraLibelle,$tranche_age.) as test,*/ 
			round(Homme,1) as NbhFinsurv,
			round(Femme,1) as NbfFinsurv 
		from T_ 
			ORDER BY put(TraLibelle,$tranche_age.); 
quit;

and the problem is that whenever someone is less than eighteen, it displays a "." instead of "Moins de 18 ans". Is it because there are only 4 under 18 women and no under 18 men?

Best regards

[KachiM]

@polpel 

If there is NONE having age < 18, you will get a missing Value which I thought it is legal value. If you want that such missing value to be handled in your way, let me know.

[polpel]

well, i have one value that is inexistent for men, but there are 4 women who are under 18.

if no one is under 18 (men or women), the "Moins de 18 ans" range shouldn't be displayed, but this isn't the case...

[ballardw]

A format approach:

proc format library=work;
value agegrp
 0 - 18		= "Moins de 18 ans"
18 <- 24		= "De 18 à 24 ans"
24 <- 29		= "De 25 à 29 ans"
29 <- 34		= "De 30 à 34 ans"
34 <- 39		= "De 35 à 39 ans"
39 <- 44		= "De 40 à 44 ans"
44 <- 49		= "De 45 à 49 ans"
49 <- 54		= "De 50 à 54 ans"
54 <- 59		= "De 55 à 59 ans"
59 <- 64		= "De 60 à 64 ans"
65 -high 	= "65 ans et plus"
;

RUN ;

proc freq data=demo;
  tables age;
  format age agegrp.;
run;

Formats have an advantage that you do not need to create additional variables, just a new format with different rules and apply in analysis, report or graphing procedure.