topic Invalid Reference Value for Proc Logistic in Statistical Procedures

Invalid Reference Value for Proc Logistic

chester2018 — Wed, 31 Jul 2024 18:23:38 GMT

Hi everyone! Hopefully you guys can help me out! Below is the code I run! I was able to run this general code for other categories such as insurance status, smoking status, etc.

proc logistic data=HHS.EPDS;
	class EPDS_Cat (ref="No");
	model guideline_levels (ref="None") = EPDS_Cat / link = glogit;
run;

But I get this error message: ERROR: Invalid reference value for EPDS_Cat. I'm trying to make the "No" as the reference group. Below is a table of the other categories.

EPDS_Cat	Frequency	Percent	Cumulative Frequency	Cumulative Percent
.	179	3.82	179	3.82
No	3896	83.19	4075	87.02
Yes	608	12.98	4683	100.00

Thanks for your help!

Re: Invalid Reference Value for Proc Logistic

SAS_Rob — Wed, 31 Jul 2024 19:05:46 GMT

I suggest that you check the formatted value itself. Likely there is a space after the o in "No". It has to match exactly. You could also just use the DESCENDING to get it to use that as the reference.

class EPDS_Cat (descending);

Re: Invalid Reference Value for Proc Logistic

Ksharp — Thu, 01 Aug 2024 02:26:58 GMT

I noticed your variable "EPDS_Cat" is numeric type.
Try using its original value like : 1 , 2

Re: Invalid Reference Value for Proc Logistic

chester2018 — Thu, 01 Aug 2024 13:54:18 GMT

Thanks! I realized that SAS made this numeric, when it should've been categorical! I made a quick change and used the value "1" as the reference group!

Re: Invalid Reference Value for Proc Logistic

ballardw — Thu, 01 Aug 2024 14:30:15 GMT

Absolutely nothing wrong with numeric values for categorical variables. "Category" is a use, not a value.

Dichotomous 1/0 valued variables can be much easier to perform some tasks with than character valued (or some odd arbitrary other number pairs).

Consider the SUM of 1/0 valued variable is the number of values=1.

The Mean would be the percentage of 1.