Large Odds Ratio

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

kea1011
Posts: 12
Joined: Sun Nov 03, 2013 5:11 pm

Large Odds Ratio

Postby kea1011 » Mon Jan 04, 2016 2:59 am

i have run a series of regressions and most of my results appear as i would expect.
however, ONE is much larger than others and larger than i am used to seeing (OR=14.53).

My main IV is a neuropsychological test score.
I am running it with control variables (age=continuous, gender=dummy, and a series of categorical dummies for education and self-reported health).
the DV is binary (yes/no) of driving status.

one of my series ended up with an educational dummy variable OR=14.53 (college education). (most of my other ORs for all variables are 3 or less!)
1) should this be a red flag???
or is this just the way things happen?

2) am i correct in stating that "the odds of observing currently driving increase 1,353% for college educated participants compared to similar participants with no education" (my reference group) ??

thank you for any guidance,
kristen
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Large Odds Ratio

Postby GerineL » Fri Jan 15, 2016 11:04 am

To me, it makes sense that education level is highly predictive of how well one does on a test. So it is not necessarily a red flag, but this is more a question about theory than about statistics.

Regarding your second question, yes you are right.
kea1011
Posts: 12
Joined: Sun Nov 03, 2013 5:11 pm

Re: Large Odds Ratio

Postby kea1011 » Fri Jan 15, 2016 4:24 pm

thank you so much for your time/response.
i have seen ORs of this number before, but when i had to WRITE OUT the 'increase in odds of over 1000 ...." that sounded suspect to me. and thought i'd better check in.
again, appreciate your thoughts!
--kea
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Large Odds Ratio

Postby GerineL » Mon Jan 18, 2016 9:24 am

Please note that while it is possible statistically to have very large odds ratios, I am not familiar with your theory nor with your data, so I cannot say with certainty whether or not something might be wrong.

Who is online

Users browsing this forum: No registered users and 3 guests

cron