Dear all,
I would appreciate your views on the following:
For Binary Logistic Regression with binary (0/1) explanatory variables :
Log (odds of event) = B0+B1*X_1 + B2*X_2 etc....
For a particular variable (say X_1) the odds ratio e^B1 tells me that someone with characteristic X1 is e^B1 more likely to have the 'event' than someone without characteristic X1 (with other variables remaining the same).
Say if we had 3 response categories: good reading skills, medium reading skills, poor reading skills and 4 binary (0/1) explanatory variables. We derive the constant, alpha1, which is associated with 'good' category and constant,alpha2, associated with 'good' or 'medium' category . I use the ordinal regression procedure in SAS so that increasing 'score': B1*X_1 + B2*X_2 +B2 X_3 + B3 X_4 means tendency towards better reading skill. Now would we say that the odds ratio e^B1 means *both* that
(i) Someone with X_1=1 has e^B1 times the odds of having good reading skill (rather than poor or medium) compared to someone with X_1=0.
AND
(ii) Someone with X_1=1 has e^B1 times the odds of having good or medium reading skill (rather than poor) compared to someone with X_1=0.
Many thanks for your views on this,
All the Best,
Kim
Dr Kim Pearce PhD, CStat
Senior Statistician
Haematological Sciences
Institute of Cellular Medicine
William Leech Building
Medical School
Newcastle University
Framlington Place
Newcastle upon Tyne
NE2 4HH
Tel: (0044) (0)191 282 0641
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|