[name],
You may be right semantically, but the "odds ratio" seems to be common usage.
Thanks,
John W.
On 9/1/2011 12:48 PM, [name] wrote:
Good morning. I have been reading your site about logistic regression and odds ratios. I don’t mean to be rude, but I think I found a key error concerning odds ratios. What is referred to there as odds ratio is actually just the odds.
http://www.appstate.edu/~whiteheadjc/service/logit/intro.htm
An odds ratio refers to a ratio of two odds. An odds ratio of 2 (See the paragraph beginning “An interpretation of the logit coefficient…”) implies that one odds is twice as large as the other odds. It does not imply anything about a probability. In fact, every probability has its corresponding odds which could then be doubled to give an odds ratio of 2.
Exactly what an odds ratio of 2 implies is still a bit vague to me. I am still trying to understand it. I think perhaps the marginal effects may be more useful, and your site has been helpful as I study marginal effects. So thank you for that.
Recent Comments