adjusted and unadjusted odds ratio

[Birdsong32]

/* Unadjusted OR from logistic regression HOSP*/
PROC LOGISTIC DATA= MB2;
CLASS VAR4(PARAM = REF REF = '1') VAR2 (PARAM = REF REF = '1')
VAR6 (PARAM = REF REF = '1') VAR10(PARAM = REF REF = '3');
MODEL VAR7(EVENT = '1') = VAR4;
RUN;

this is the unadjusted odds ratio code im running

its saying invalid reference value for VAR4  

[Reeza]

VAR4(PARAM = REF REF = '1')

It means you don't have a value of 1 for your Var4.

Run the following to see the values. I can't recall if it should be the formatted or unformatted value you include in the REF - the documentation does state which though.

proc freq data=mb2;
table var4;
run;

---

@Birdsong32 wrote:

/* Unadjusted OR from logistic regression HOSP*/
PROC LOGISTIC DATA= MB2;
CLASS VAR4(PARAM = REF REF = '1') VAR2 (PARAM = REF REF = '1')
VAR6 (PARAM = REF REF = '1') VAR10(PARAM = REF REF = '3');
MODEL VAR7(EVENT = '1') = VAR4;
RUN;

this is the unadjusted odds ratio code im running

its saying invalid reference value for VAR4  

---

[Birdsong32]

Should i change my variable name?

if so, how do i go by doing so?

[Reeza]

No, it's the value of '1' SAS that's the issue. SAS doesn't find any values that are 1 to be the reference. This could be either because you have no 1s in your data set or because you have a format applied to the VAR4 variable that displays the 1s as "Yes" for example. If the variable has a format, REF expects the formatted value, not the unformatted value.

But we're guessing a bit. The log and the results from the code I included would allow us to definitively answer your question.

[StatDave]

If VAR4 has an associated format, then you need to specify the formatted value. And I strongly advise against including variables in the CLASS statement that are not also used in the MODEL statement. Any observation with a missing value on any of those unused variables will be omitted from the model fitting process, so you could end up using fewer observations than possible.