Problem with ARRAY

[deleted_user]

Hi everybody! i'am new here, sorry if i am not very clear!
I have a problem with ARRAY:
i have data with firms identifiers, salaries and number of staff for each trimester of each year since 2000 to 2005.
SO I want to create a new table wich gives to me a unique line for each firm and in each i would have the salary by trimester per year (like 2000T1 2000T2 2000T3...2000T5) so i have to do this for the salary per trimester year and the number of staff per trimester year...
But i cant use ARRAY AND RETAIN,
if anybody has an idea, i thanks him !

[deleted_user]

Faced with doing this for multiple variables, I would code it by arrays because I can. But if you can't, then you may not have any confidence your solution is correct, and I would suggest you leave arrays alone.

Instead, look up the Factual Material on the Transpose Procedure. There are quite a few very well documented examples for rearranging rows of data into columns.

As a hint, you will have to transpose once each for salary and staff count, and then match merge the results together.

You might want to be very careful about your time line too. I thought a trimester was a three part period, with the third trimester being the last. That being so, I don't see how 2000 can have five trimesters. Perhaps it is a mistake, but it is mistakes like that, which would kill your use of an array.

Kind regards

David

[deleted_user]

yes it was a mistake, i want to create variables trimester per year like that:
2000T1 2000T2 2000T3 2000T4.....20005T4

[deleted_user]

i post my email adress: if anybody could help me i'll give him my programme and i could send him what really i have to do!
vascogil_1@hotmail.com

[deleted_user]

hi
can you be little more specific about your data structure like columns and rows or can u write sample dataset for which u want solution.
because untill have an idia about data its not easy to give solution.
thanks

[deleted_user]

it looks like that:
the columns: variables:
Ident/Salary/staff/year/trimester
x1 /1200 /2 /2000/1
x1 /2100 /4 /2000/2
x1 /3440 /6 /2000/3
...for each ident as x we have observations for salary and staff per year trimester...
until
x1 /2300/ 6 /2005/4

etc and we have many ident from x1 to Xn...

so i want to create a new table which gives me
Only one ident per line but all staff and salaries per each year trimester

like:
ident/salary_2000T1/salary_2000T2/.../salary_2005T4/staff_2000T1/staff_2000T2 etc...
at least i have to create a new table with 4 trimesters per each of my 6 years but two times one for the staff an other for the salaries so 48 variables more ident and a anothre which doesnt matter.

thats all!

[deleted_user]

If you haven't installed the SAS Online documentation, then visit the web based documentation at http://support.sas.com/onlinedoc/913/docMainpage.jsp and look up Base SAS / Transpose Procedure.

Observe the example using the BY statement and how that relates to your IDENT variable.

Now observe the VAR statement and how that relates to your SALARY and STAFF variables.

Finally, observe that the ID statement will identify the variable that names the transposed columns.

Note that the PREFIX statement will identify the prefix to place before the value of your ID statement to name the column. So, with a prefix value of SAL and an ID of PERIOD you will have columns named SAL20001. For naming purposes you are better off constructing a variable called PERIOD that concatenates YEAR and "Trimester" into a value that does not contain characters you cannot use in variable naming.

Do the transposition first for SALARY and then for STAFF so that you can set the prefix appropriately for each VAR as you transpose it.

Then match merge the two tables together to give you a single record for each Ident value. Message was edited by: dkvj

[deleted_user]

okay, i understood the problem, but i have really to do this with ARRAY and RETAIN.
I will try to TRANSPOSE the data table and i will see what it does. But i have to finish with array et retain!
I think i have to use it because there are too many variables to create.

They give to me this example, so i have to use it for my data...

data MANYPER;
input ID TIME SCORE;
datalines;
01 1 3
01 2 4
01 3 5
02 1 7
02 3 9
03 1 6
03 2 5
03 3 4
;
run;
proc sort data=MANYPER; by ID TIME; run;
data ONEPER_REC;
array S[3] S1-S3;
retain S1-S3;
set MANYPER;
by ID;
If first.id then do i=1 to 3;
S =.;
end;
S[TIME]=SCORE;
If last.id then output;
keep id S1-S3;
run;

[deleted_user]

the problem with that program is that mine have more conditions:
i have trimesters (4) per years (6) and i have to build a data which gives me salary and the staff for every variable...
i really dont know how to do...

[1162]

Try this solution which does not use array:

data in;
infile cards dlm='/';
input Ident :$8. (Salary staff year trimester) (:8.);
Period = compress(Year||"T"||Trimester);
cards;
x1 /1200 /2 /2000/1
x1 /2100 /4 /2000/2
x1 /3440 /6 /2000/3
x2 /1500 /1 /2000/1
x2 /2900 /3 /2000/2
x2 /3140 /5 /2000/3
;
run;

proc sort data=in; by ident period; run;

proc transpose data=in out=tmp1 (drop=_name_) prefix=salary_;
by ident;
id Period;
var salary;
run;

proc transpose data=in out=tmp2 (drop=_name_) prefix=staff_;
by ident;
id Period;
var staff;
run;

data out;
merge tmp1 tmp2;
by ident;
run;


The output looks like this:

Ident salary_2000T1 salary_2000T2 salary_2000T3 staff_2000T1 staff_2000T2 staff_2000T3

x1 1200 2100 3440 2 4 6
x2 1500 2900 3140 1 3 5

[Matthew_Galati]

Is this question related to Mathematical Optimization and Operations Research with SAS? If not, this is the wrong forum.

[deleted_user]

Thank you so much for your help!
It's a way to solve the problem, but my "boss" want I to do with array and retain!
so if anybody knows, he or she is welcome!

[deleted_user]

This the solution:



data projet.etablissements5;
array salary[24] salary1-salary24;
array staff[24] staff1-staff24;
retain salary1-salary24
staff1-staff24;
set projet.etablissements4;
by ident;
if first.ident then do i=1 to 24;
salary=.;
staff=.;
end;
salary[periode]=Masse_sal;
staff[periode]=Effectif;
if last.ident then output;
keep ident APE salary1-salary24 staff1-staff24;
run;