topic Using If Statement to Create Variable in SAS Programming

Using If Statement to Create Variable

leackell13 — Fri, 17 Dec 2021 21:21:05 GMT

Hi,

I have a character variable, SmokingStatus, with these values: "Never Smoker","Former Smoker","Light Smoker", and "Heavy Smoker". I am trying to assign them numbers so I can interpret them easier in a regression:

DATA SubQ.dataclean3;

	LENGTH SmokingStatus $22.
			Smoking 8.;
	
	SET SubQ.combined;
		
		IF SmokingStatus="*Unknown" THEN Smoking_Status=' '; /* Highlight missing values */
		IF SmokingStatus="Never Smoker" THEN Smoking=0; /* Change smoking group to easier variable to use in regression */
		IF SmokingStatus="Former Smoker" THEN Smoking=1;
		IF Smoking_Status="Light Smoker" THEN Smoking=2;
		IF Smoking_Status="Heavy Smoker" THEN Smoking=3;
	
	RUN;

The problem is they assign 0 and 1 correctly but not 2 and 3; for light and heavy smokers, Smoking is shown as missing. Does anyone know why this is? I have also tried this making Smoking a character variable with $22. length and quotations around the number values.

Thanks.

Re: Using If Statement to Create Variable

yabwon — Fri, 17 Dec 2021 21:41:47 GMT

IF SmokingStatus="*Unknown" THEN Smoking_Status=.;

Numeric missing is . (a dot)

{EDIT:} sorry didn't spot the issue. You are missing _ in the variable name:

SmokingStatus is not the same as Smoking_Status

{EDIT2:} The length statement should be like:

LENGTH SmokingStatus $ 22 Smoking 8;

Bart

Re: Using If Statement to Create Variable

Kurt_Bremser — Fri, 17 Dec 2021 21:43:08 GMT

The spelling must match exactly, including any non-printable characters that might be contained in the string. Print the string with a $HEX format to reveal such characters.

Re: Using If Statement to Create Variable

Reeza — Fri, 17 Dec 2021 21:51:18 GMT

@leackell13 wrote:

Hi,

I have a character variable, SmokingStatus, with these values: "Never Smoker","Former Smoker","Light Smoker", and "Heavy Smoker". I am trying to assign them numbers so I can interpret them easier in a regression:

How do you think that coding these to numbers will make it easier to interpret your results? They're categorical variables so need to be included in a CLASS statement and then the output would be labeled as "Never Smoker" vs "Heavy Smoker" compared to 1 vs 3? The first version would be much easier to read.

Are you planning to treat these as a continuous ordinal variable instead of categorical?

@leackell13 wrote:

Hi,

I have a character variable, SmokingStatus, with these values: "Never Smoker","Former Smoker","Light Smoker", and "Heavy Smoker". I am trying to assign them numbers so I can interpret them easier in a regression:
DATA SubQ.dataclean3;

	LENGTH SmokingStatus $22.
			Smoking 8.;
	
	SET SubQ.combined;
		
		IF SmokingStatus="*Unknown" THEN Smoking_Status=' '; /* Highlight missing values */
		IF SmokingStatus="Never Smoker" THEN Smoking=0; /* Change smoking group to easier variable to use in regression */
		IF SmokingStatus="Former Smoker" THEN Smoking=1;
		IF Smoking_Status="Light Smoker" THEN Smoking=2;
		IF Smoking_Status="Heavy Smoker" THEN Smoking=3;
	
	RUN;
The problem is they assign 0 and 1 correctly but not 2 and 3; for light and heavy smokers, Smoking is shown as missing. Does anyone know why this is? I have also tried this making Smoking a character variable with $22. length and quotations around the number values.

Thanks.

Re: Using If Statement to Create Variable

leackell13 — Fri, 17 Dec 2021 21:57:55 GMT

That is a good point; I think I was afraid that the values would be too large. So under the CLASS statement, I can just put SmokingStatus (REF='Never Smoker')?

Re: Using If Statement to Create Variable

Reeza — Fri, 17 Dec 2021 22:03:43 GMT

@leackell13 wrote:
That is a good point; I think I was afraid that the values would be too large. So under the CLASS statement, I can just put SmokingStatus (REF='Never Smoker')?

Yes, that is how you'd specify a reference level.

I would also add PARAM=REF to indicate that you're using the One hot encoding/dummy variables/Referential parameterization instead of GLM which is the SAS default.