Changing values of a macro-variable (as a boundary)

[GuiVtzl]

Hello everyone,

I have a little problem. I have this kind of data set with say, 1 000 000 observations :

VAR_1 VAR_2 VAR_3 [...] VAR_122  SPAN

The SPAN variable gives me a number between 1 and 122 and what I'd like to do is count for each observation the number of missing values between VAR_1 and VAR_&SPAN.

I don't really know how to perform that calculation since it's a "row calculation". Writing it wrong, it would look like

N_MISS = count missing(VAR_1-VAR_&SPAN)

But I have no idea about how to perform that calculation for each row,    I only know how to do that in column (more simple ahah)

Could you help me please ?

Thanks a lot

[jakarman]

use array-s Base SAS(R) 9.4 Procedures Guide, Second Edition.

[GuiVtzl]

Thank you Jaap for your answer.

I am not that familiar with arrays, but I tried looking at that.

So I guess I should start doing something like :

array var

var_1-var_122;

array span

SPAN;

array n_miss

n_miss;

Is that correct ?

But then i don't see how i should do my loop.

I thought it would be :

do  i=1 to 122

n_miss = cmiss (of var[1]-var[SPAN];

leave;

end;

run;

But again, I am not at ease yet with arrays.

Could you correct me please ?

Thanks a lot

[RW9]

I think you would need something more like:

n_miss=0;

do I=1 to 122;

     if var=. then n_miss=n_miss+1;

end;

[GuiVtzl]

Thank you RW9,

the problem is that your code seems to count the missing values for the 122 variables "VAR" for each observation.

In my case I need for each obs to count the missing values from VAR_1 to VAR_j where j is the value of variable SPAN.

[RW9]

Maybe I misunderstand, but for the code above you would get:

datastep start.

arrive at observation1.

set n_miss to 0

start do loop with 1

if var[1] is missing the n_miss+1;

do loop with 2

if var[2] is missing then n_miss+1;

...

end loop, n_miss contains, for the first observation, a count of all columns var1-var122 where the value is missing.

now move onto observation 2.

This seems to be what you are asking?

[jakarman]

My fault Id did not include a good link for array-s, this one is better. SAS(R) 9.4 Language Reference: Concepts, Second Edition fo concepts and as reference SAS(R) 9.4 Statements: Reference, Second Edition  your array declaration should be something like:

array aspan{*} var_1 - var_122 ;  /* the * is counting the 122 automatically */

After that you can refer to the element of an array using loop-processing.

This is offering full freedom like matrix processing.

      do i=1 to ..;

         span= <some processing  aspan{i} > ;

      end;

Reviewing your question, There are a lot of functions that are working on a list of variables. My first link was one in that direction.
you are just wanting to count the number of missings? SAS(R) 9.4 Functions and CALL Routines: Reference, Second Edition (NMISS)
span=nmiss(of var_1 - var_122) ;

That is using the list processing of variables SAS(R) 9.4 Language Reference: Concepts, Second Edition.
That a list of variables can be the argument is not a very common concept in programming languages.

[GuiVtzl]

Thank you both !

I am going to put a little example to show what I want to count (in bold the variables I already have, I want to create the variable N_MISS):

VAR_1     VAR_2      VAR_3      VAR_4      VAR_5     SPAN     N_MISS

     5               10           .               .                    10          5               2

     2               3             .               .                    .               3               1

     .               .               1               5                    .               4               2

------------------------------------------------

I am sorry if I get it wrong but I have the impression that when writing things like nmiss(of var_1 - var_122) ; or                  

n_miss=0;

do I=1 to 122;

     if var=. then n_miss=n_miss+1;

end;

I will get N_MISS values 2 3 3 instead of the 2 1 2 above

My problem seems to me (again, I am probably wrong !) that my loop isn't fixed since for each observation (row) I don't use the same span to count. For my example above, 1st line I count on var1-var5, 2nd line I count on var1-var3 and 3rd line I count on var1-var4.

Thank you very much for your help, those arrays are totally new to me (discovered them this morning ahah)

[RW9]

Ok, so you want to count the missing based on another variable span, that's fine just change to:

n_miss=0;

do I=1 to 122;

     if var=. and I <= span then n_miss=n_miss+1;

end;

[Kurt_Bremser]

I think this would be easier and more intuitive:

do I=1 to span;

     if var=. then n_miss=n_miss+1;

end;

[GuiVtzl]

I mixed all of your suggestions and it did exactly what I wanted !

Thanks a lot guys for your help ! Plus, it made me discover the arrays ! Great !

Have a nice day !

[andreas_lds]

Is it really necessary to do the iteration?

data work.want;

   set work.have;

   length n_miss 8;

   array vars

var_:;

   n_miss = nmiss(of var_:) - (dim(vars) - span);

run;

The result of nmiss(of var_:) has to be reduced by the number of variables which are expected to missing.

[Kurt_Bremser]

If it is not guaranteed that variables beyond SPAN are always missing, this approach will give wrong results.

[And]

Hello,

Here is a macro which also provides the desired result.

%macro tocalc (nbvar);

data want;

set have;

mc = span-(%eval(&nbvar) - N(of var1-var&nbvar));

%put var&nbvar;

run;

%mend tocalc;

%tocalc(5);