I am conducting a multinomial logistic regression to test the effect of socio demographic variables (age, gender etc) in the choices of different chocolates. The choices are categorical (choice 1, choice 2 or choice 3). I have found that one variable affect the choices significantly (compared to the base category) and the odds ratio is very high (80000). Is it normal to have very high odds ratio?
Hello Tarapada,
Chocolate selections, what an interesting study! I presume you're referring to the exponentiated value of the regression coefficient, which quantifies the expected change in odds of the target outcome occurring, given a unit change in the specific IV.
I can't recall having seen a (exp(B)) value as high as 80,000 (implying a regression coefficient of about 11.29) but that doesn't mean it couldn't be true. I suppose the only concern might be if somehow an IV was a direct proxy for the DV. It doesn't sound as if that's the case with your data set.
Good luck with your work!
Hello Tarapada. Can you show us a frequency distribution for the outcome variable? I am wondering if one of the 3 categories has almost no observations in it.
Which variable has the high odds ratio? If it is categorical, can you show us a cross-tabulation of that variable with the outcome variable too? HTH.
- Badges
- Science topic
- Similar topics
- Computer Science
- Research Topics