topic Re: Why there is a letter "I" in numeric variable of SAS dataset? in New SAS User

Why there is a letter "I" in numeric variable of SAS dataset?

kumarayas — Thu, 26 Mar 2020 12:46:53 GMT

I have a SAS dataset where I see a letter "I" (Capital i) in numeric column. What does it mean?

I don't understand as it is a numeric variable and it should not contain this unless it means something else. I tried to add/subtract a number to it and it returned "." (null).

See below picture:

Here, "Factor" variable in 2nd row has letter "I" as Highlighted. The 3rd variable "db" is same as "factor" but with no format applied.

Table properties:

Compressed: CHAR

Encoding: wlatin1 Western (windows)

SAS version: 9.4m5

Re: Why there is a letter "I" in numeric variable of SAS dataset?

Tom — Thu, 26 Mar 2020 12:50:33 GMT

Most likely it is a special missing value. In addition to the normal missing value, which is represented by a period, SAS allows 27 special missing values. They are represented in code by period followed by a letter or underscore. In listing/printing the period is not displayed.

Try something like this to see how many observations have .I for the DB variable.

proc print data=MYDATA;
  where db=.i ;
run;

Re: Why there is a letter "I" in numeric variable of SAS dataset?

Kurt_Bremser — Thu, 26 Mar 2020 12:52:50 GMT

"Special Missing" values are created like this:

data test;
db = .I;
run;

Re: Why there is a letter "I" in numeric variable of SAS dataset?

FreelanceReinh — Thu, 26 Mar 2020 15:14:42 GMT

Hi @kumarayas,

While programmers are free to use special missing values such as .I for arbitrary purposes, there are contexts in which .I has a special meaning: infinity. For example, the DIVIDE function and PROC TTEST produce this particular special missing value with the meaning "infinity" in certain situations (see code examples below). But even then it's still a missing value and recognized as such, e.g., by functions like MISSING, NMISS or CMISS.

/* Example 1 (Result: x=.I) */

data test;
x=divide(1,0);
run;

/* Example 2 (Result: UpperCLMean=.I) */

ods output conflimits=cl;
proc ttest data=sashelp.class sides=u;
var age;
run;

Re: Why there is a letter "I" in numeric variable of SAS dataset?

novinosrin — Thu, 26 Mar 2020 15:40:48 GMT

Thank you Sir @FreelanceReinh for sharing. Always spot on and to the point. Kudos! Interestingly, I ran into something similar a week ago viz. special missings, though was able to figure the details after having read an amazing paper by @Quentin on special missings for BY GROUP processing that I used for scoring. Cheers!

Re: Why there is a letter "I" in numeric variable of SAS dataset?

FreelanceReinh — Thu, 26 Mar 2020 16:16:30 GMT

@novinosrin wrote:

... read an amazing paper by @Quentin on special missings ...

I assume you mean "How to Represent Missing Data: Special Missing Values vs. 999999999." Indeed, this is really good and timeless. Thanks.

Re: Why there is a letter "I" in numeric variable of SAS dataset?

Quentin — Thu, 26 Mar 2020 19:00:06 GMT

Thanks for your kind words about my paper, @novinosrin and @FreelanceReinh . That was the first UG paper I ever wrote. (I was too scared to present, so I did it as a poster for long-lost NESUG). Always nice to know that a paper has been read!