topic Re: How to count all the miss values for an entire observation, taking into consideration the 0 in SAS Programming

How to count all the miss values for an entire observation, taking into consideration the 0

Feragon42 — Tue, 10 Jan 2017 15:52:27 GMT

Hello everybody,

Today I come with this problem that I hope you can help me.

Let's say that I have this dataset:

data have;
input A $ B C D $;
datalines;
try 0 57 0
do . 42 something
with 442 . this
;

And I need to return this:

The new variable, N_MISS, count how many missing character I have in every observation. I know can do this easily with the function CMISS(OF A-D), but the problem is that I need to consider the 0 as a missing value.

I create a format to identify with 0 and 1 when the value are missing or not, and then sum al the variables of the observation and have the amount of not missing values.

PROC FORMAT LIBRARY = FORMATS;
	INVALUE $CMISS_COUNT '0' = 0
	                     ' ' = 0
		             '.' = 0
			     OTHER = 1;
	INVALUE NMISS_COUNT 0 = 0
			    . = 0
			    OTHER = 1;
RUN;

But, when I apply a format, the value of the character still the same for the sum. I can use the fuction PUT(A, $CMISS_COUNT.), but in the real excercise in don't have only 4 variables but 250 ... aaaaand do that 250 times is... well.

I hope I can express myself, sorry for my newbie english.

Goodbye, and thanks for all.

Re: How to count all the miss values for an entire observation, taking into consideration the 0

ballardw — Tue, 10 Jan 2017 16:06:46 GMT

You can use the translate function in cmiss but you won't be able to use the "of" notation. Or you create a separate array of temporary variables that you apply translate to.

data junk;
   x = '0';
   y = '123';
   z = ' ';
   nmiss = cmiss(translate(x,' ','0'),y,z);
run;

This approach does have a potential issue in that it will treat '0000' as missing as well. If that is not desired and the value appears in your data then you may be better off creating an array of temporary variables with some logic involving If/then for exactly '0'.

Re: How to count all the miss values for an entire observation, taking into consideration the 0

collinelliot — Tue, 10 Jan 2017 16:24:40 GMT

Something like this?


data want;
    set have;
    array nvars{*} _numeric_;
    array cvars{*} _character_;
    do i = 1 to dim(nvars);
        n_miss = sum(n_miss, (nvars(i) in (., 0)));
    end;
    do j = 1 to dim(cvars);
        c_miss = sum(c_miss, (cvars(j) in ('.', '0')));
    end;
    all_miss = sum(n_miss, c_miss);
run;

Re: How to count all the miss values for an entire observation, taking into consideration the 0

RW9 — Tue, 10 Jan 2017 17:18:15 GMT

I can't test this right at the moment, but what about something like:

data want;
  set have;
  total=countc(cats(of a--d),'.0');
run;

Re: How to count all the miss values for an entire observation, taking into consideration the 0

Feragon42 — Tue, 10 Jan 2017 18:11:20 GMT

Excelent, this works. Very thanks.

Re: How to count all the miss values for an entire observation, taking into consideration the 0

Feragon42 — Tue, 10 Jan 2017 18:13:18 GMT

Interesting method.... But, when the database present a value like '200', the function will detect 2 extra missings thanks to the 0.