In logistic regression, does an odds ratio of zero make sense, and if so, what's the interpretation.
I've only been able to locate one reference that specifically refers to OR of 0: Making Sense of Data: A Self Instruction Manual on the Interpretation of Epidemiological Data (Abramson et al, 2001 ) p113 which just states that an odds ratio of zero indicates one of the odds being compared must be zero, and that it indicates a strong negative association (unless the other odds is close to zero).
If an OR of 1 indicates no association, OR of 0 must indicate infinite association ? In that case, is a predictor with an OR of zero a predictor that results in complete separation?
I'm not sure the background of the study is relevant to the question, but just in case: the DV is the presence or absence of medium scale residential development in a neighbourhood, and the IV with a odds ratio of 0 is the median house / land sales price per square metre, which theory says is likely to be a predictor of the scale of residential development. The variable was significant in the logistic regression.