topic Re: Recoding missing variables --seemingly simple program not working in SAS Programming

Recoding missing variables --seemingly simple program not working

kwhood — Fri, 08 Jul 2016 20:49:37 GMT

I have a categorical variable that has either 1) missing "." data or a category for 2) "CAH" or 3) "children's". I need to redesignation the Missing or blank observation should really have been coded as "adult" for an adult care hospital.

when I ran a frequency it told me how many hospitals I had in CAHs or Children's hospitals;

So I tried to recode the variable with the following statement:

if type="." then type="adult";
else if type="CAH" then type="CAH";
else if type="Children's" then type="Children";
run;

proc freq data=weighting;
tables type;
run;

I received the same output as posted above. I also tried to having the missing outside if quotes (see below)

if type=. then type="adult";
else if type="CAH" then type="CAH";
else if type="Children's" then type="Children";
run;

When I did this I got this result:

So all the 3 catergories became recoded as "adult" hospitals. I don't know what I am doing wrong. Can someone help me fix the program so that the missing data become adult hospitals and the other existing categories stay the same??

Thank you!!!!

Re: Recoding missing variables --seemingly simple program not working

ballardw — Fri, 08 Jul 2016 21:29:32 GMT

Missing is a bit different for character and numeric. the . is for numeric.

Try

if missing(type) then type='Adult';

if missing='' then type='Adult'; /* that is two single quotes with no space between*/

Re: Recoding missing variables --seemingly simple program not working

stat_sas — Fri, 08 Jul 2016 21:29:41 GMT

Hi,

Instead of recoding type in output, try to recode type in orginal data set weighting then run proc freq to get the desired results.

Re: Recoding missing variables --seemingly simple program not working

Tom — Sat, 09 Jul 2016 01:08:13 GMT

One of those two data steps should have given messages about needing to convert between numeric and character data.

If TYPE is numeric then you can use TYPE=. to test for missing.

But if TYPE is character then you need to use TYPE=' ' to test for missing.

You can also use MISSING(TYPE) to test for missing.

Re: Recoding missing variables --seemingly simple program not working

kwhood — Sat, 09 Jul 2016 01:53:28 GMT

Super! Thanks so much. I didn't know that at all.