No such column in SAS dataset but able to run the code

[ashiah]

Hello everyone,

I'm relatively new to SAS, having previously worked extensively with R. Currently, I'm using SAS during my co-op placement. While reviewing the code written by the previous co-op student, I came across something that confused me.

The SAS code contains a statement like this

"if cough =1 or wheeze=1 or ... =1, then symp=1".

However, I couldn't find any columns with names like cough, wheeze, or others mentioned in the SAS dataset that was read.

In R, it will only work/run when it has the column named as this. Could anyone help me with this? 

[Kurt_Bremser]

Maxim 2: Read the Log.

If a variable is used which is not in an incoming dataset, or assigned a value somewhere in the code, an "uninitialized" NOTE is written to the log.

[ashiah]

Hi Kurt,

If a variable is used which is not in an incoming dataset, but the code could run. There is no error in SAS log.

Thanks.

Chloe

[ashiah]

Sorry, I mean the variables are used not in my read in the dataset. At least, no column name matched.

[Reeza]

We can't answer why that occurs, for that answer you'll need to check with the previous student. Or at minimum post the full code.

It could be as simple as an incorrect version of the coop students code or bad programming.

[ashiah]

Hi Reeza,

Due to confidentiality, I revised part of the variable names. But the partial code would look like that,

libnmae data1 "xxxx\xxxxx\xxxxx\"; /*Path*/

data data1;

          set data1.xxxxxxxxx;

run;

data data2;

          set data1;

          if cough =1 or phlegm =1 or wheeze =1 then coughsymp =1; else coughsymp =0;

          if xxxx =1 or xxxx= 1 or xxxx =1 then zzzz =1; else zzzz=0;

run;

But after running the code, there is no error in the log and no such column in the read-in dataset. Therefore, I got confused.

[Reeza]

Did you run a proc contents on your input data set?

proc contents data=data1.xxxxx;
run;

EDIT: From this, look at the variable name and label. It's possible the variable has a label so you're not seeing the variable name but the label instead. 

[Tom]

Your won't get an ERROR message since it is not an error (other than a logic error).

But if the variables really do not exist then you should get notes. 

Example:

271  data have;
272   id=1;
273  run;

NOTE: The data set WORK.HAVE has 1 observations and 1 variables.
NOTE: DATA statement used (Total process time):
      real time           0.01 seconds
      cpu time            0.00 seconds


274
275  data want;
276    set have;
277    if cough =1 or wheeze=1 then symp=1;
278  run;

NOTE: Variable cough is uninitialized.
NOTE: Variable wheeze is uninitialized.
NOTE: There were 1 observations read from the data set WORK.HAVE.
NOTE: The data set WORK.WANT has 1 observations and 4 variables.
NOTE: DATA statement used (Total process time):
      real time           0.01 seconds
      cpu time            0.00 seconds

But it is possible to write a step where there is no obvious place where COUGH is assigned a value, but the uninitialized note does not appear.  The most obvious it to include the variable in an array.

So in this example COUGH is included in the array so it no longer generates the note.

286  data want;
287    set have;
288    array x cough ;
289    if cough =1 or wheeze=1 then symp=1;
290  run;

NOTE: Variable wheeze is uninitialized.
NOTE: There were 1 observations read from the data set WORK.HAVE.
NOTE: The data set WORK.WANT has 1 observations and 4 variables.
NOTE: DATA statement used (Total process time):
      real time           0.00 seconds
      cpu time            0.00 seconds

[Kurt_Bremser]

---

@ashiah wrote:

Hi Kurt,

If a variable is used which is not in an incoming dataset, but the code could run. There is no error in SAS log.

Thanks.

Chloe

---

If you do not get the "uninitialized" NOTE, then the variable is either read from a dataset (or an external file with an INPUT statement) or is assigned a value somewhere in the DATA step.

[Reeza]

Is there a KEEP/DROP statement somewhere in the data step?

[Tom]

---

@ashiah wrote:

Hi Kurt,

If a variable is used which is not in an incoming dataset, but the code could run. There is no error in SAS log.

Thanks.

Chloe

---

That is because it is NOT a syntax error.   SAS is designed to allow you to reference variables that you have not yet defined.  In that case the data step compiler will create the variable and make its best guess at whether you wanted a numeric or a character variable and if character what length it should have.

For a more complete answer post the complete log for the data step in question.

[Reeza]

Because SAS uses = as both assignment and equivalence it does work differently.

If it finds a variable that doesn't exist, it will create it and initialize it as missing. This can be useful in some circumstances, just a difference in how the defaults are handled.

FYI _ I would double check the data was imported or created properly in previous steps if data is missing or if it's pointing to the correct data step. In a case like this, I'd expect to see an error prior to this step most likely.

[ballardw]

SAS as a language will go a bit to help prevent code from failing from, hopefully, minor errors.

One of those is creating variables when the code mentions them if they are not previously defined such as appearing in a source data set, or statements that are intended to create variables such as Length, Array, informat or format and label.

Even misspelling a variable will create one with the new spelling. just in case you meant to and just haven't provided the rest of the code. SAS generally does not throw an error but uses the new variable with a missing value. {Implied hint: long complex names are easier to misspell}

Another thing I would be concerned with inherited code is tracing everything back to when the data is read into SAS.

A common occurrence with new SAS users is a reliance on Proc Import or menus that basically invoke Import to read data.  New files read with old code using Imporat can have different variable names because the column headings in the source file changed. So a variable that was previously "Cough" might become "Heavy_Cough" or something similar without actually changing the meaning of the variable just the name.