topic Re: recoding variables in New SAS User

Re: recoding variables

Eric1337 — Wed, 26 Jan 2022 22:38:51 GMT

I am trying to create a variable that is dependent on responses to 3 other variables that is grouped under one main category in a survey.
If the respondent has responded yes ( which is coded as 1) to more than one of the three variables then , then this new variable is coded as 1
Or in other words the respondent must have responded yes to more than one of the three variables
Since all the three variables are numeric, I made a new variable in a data step that calculates the sum of these responses to the three variables .
If the sum is greater than 1 then this new variable is coded as 1 otherwise 0 .
This is just to determine the percent of this new variable . Just descriptive statistics
Wondering if anyone has any thoughts?
Thanks

Re: recoding variables

PaigeMiller — Wed, 26 Jan 2022 23:07:37 GMT

Simple math. Add up the zeros and ones, if it is 2 or greater then new variable is 1.

Percent of 0/1 binary variables that are 1: use PROC SUMMARY and compute the mean of this new variable.

Re: recoding variables

ballardw — Wed, 26 Jan 2022 23:54:05 GMT

Recode = ( sum(var1, var2, var3) > 1);

SAS will return 1/0 for true/false of comparisons. If the variables are all numeric coded 1/0 then sum is the number of 1 responses.

Caveat: the above will return 0 for the comparison if all the responses are missing. If you require that at least some number of the variables have an answer, such as 2 for example, then perhaps:

If n(var1,var2,var3)>2 then Recode = ( sum(var1, var2, var3) > 1);

Which will only assign a value to Recode when at least two of the variables have values.

If your variables have a common "root" name you can use short cut lists like sum(of var: ) The : says to use all of the variables whose names start with VAR (caution).

Re: recoding variables

Eric1337 — Wed, 26 Jan 2022 23:57:35 GMT

Thanks you Paige