05-09-2017 03:39 PM
I am using SAS 9.4 to run a logistic regression predicting 5% weight loss (fiverper, 0 =No, 1 =yes) with time overweight (continuous variables) as the independent variable -
This is the first way I tried it -
proc logistic descending;
class Fiveper (param=ref ref='1');
model time= now weight bmi raceth0_1 raceth0_2 college status level binge;
** QUESTION** Do I need to class my other dichotomous variables - college, status, level, binge - or are they ref =0 automatically?
This is the second way I tried it -
proc logistic data =dummy_11;
class fiveper (ref = "1") raceth (ref = "0")college (ref="1") status (ref="1") level (ref="1") binge (ref="1") / param = ref;
model fiveper = now weight bmi raceth college status level binge / link = glogit;
On the second way, I feel like it is not reference correctly, even though I put college (ref="1') it is coming up in the odds ratios tables as college 0 vs 1.
Thank you for your help!
05-09-2017 04:14 PM
You set the reference level equal to 0. So the odds ratio will compare all other values to the reference level, 1. The 'other' values are 0, therefore the odds ratio becomes 0 vs 1.
The second method is correct, especially if you're unsure of interpretations.