topic Re: Proc Logistic help in Statistical Procedures

Proc Logistic help

shortyofhb — Thu, 17 Dec 2020 08:04:27 GMT

Hello all,

I am trying to use a Proc Logistic to run a regression, but am getting stuck.

I keep getting a note in the log that says:

NOTE: The REF= option for the response variable is ignored.
NOTE: PROC LOGISTIC is fitting the cumulative logit model. The probabilities modeled are summed over the responses having the lower
Ordered Values in the Response Profile table. Use the response variable option DESCENDING if you want to reverse the
assignment of Ordered Values to the response levels.
NOTE: Convergence criterion (GCONV=1E-8) satisfied.

I know that when running a proc logistic the outcome variable should be dichotomous right? I believe the issue is that I have values coded for No (1) and Yes (2), but also have 9 set to missings, I see in my Proc Logisitc output its reading the 9, how would I make it not read this extra value and only the 1 and 2?

Re: Proc Logistic help

Kurt_Bremser — Thu, 17 Dec 2020 09:21:18 GMT

Use a WHERE= dataset option to exclude certain observations.

Re: Proc Logistic help

shortyofhb — Thu, 17 Dec 2020 09:34:27 GMT

So I would have

Proc data=x;

Class Gender (Param=Ref Ref="Male") Outcome (Param=Ref Ref="No");

Where Outcome = "No", "Yes";

Model Outcome=Gender;

run;

Or how would I use the where to only read "No" and "Yes" but not the 9 that codes for missing?

Re: Proc Logistic help

Kurt_Bremser — Thu, 17 Dec 2020 09:42:05 GMT

My personal preference is the dataset option:

proc logistic data=x (where=(Outcome in ("No","Yes")));

Note that

Outcome = "No", "Yes"

is invalid comparison syntax anywhere in the SAS language.

Re: Proc Logistic help

shortyofhb — Thu, 17 Dec 2020 09:48:59 GMT

Hello when I tried it the way you suggested I get an error in the log:

ERROR: WHERE clause operator requires compatible variables.

Note: I must use the Param=Ref Ref="Male" ... etc line

Re: Proc Logistic help

Kurt_Bremser — Thu, 17 Dec 2020 10:09:16 GMT

In the WHERE statement and WHERE= dataset option, you need to use the raw values stored in the dataset, not the formatted ones.

Maxim 3: Know Your Data; run a PROC CONTENTS to see variable attributes (including assigned formats), and then look at the formats to see which raw values are expected.

Re: Proc Logistic help

shortyofhb — Thu, 17 Dec 2020 10:15:36 GMT

Ah okay maybe I didn't ask the question clearly.

So I need to run a logistic regression model using the parameterization method. I know my outcome variable of interest should be Yes=1 and No=2. (These are my newly formatted variables however, I used an array to recode No=0 to No=2)

In my data set, for my outcome variable of interest, I have Yes=1, No=2, and Missing=9. If I run a proc freq, I have 1,2, and 9 pop up. Is there a way for my to run this proc logistics while ignoring the 9?

Re: Proc Logistic help

Kurt_Bremser — Thu, 17 Dec 2020 10:20:06 GMT

where=(variable ne 9)

as dataset option.

Re: Proc Logistic help

shortyofhb — Thu, 17 Dec 2020 10:21:48 GMT

Great that worked, thank you!