Logistic Regression with Continuous Predictor Variable

[Rekansha]

Hi,

Just a quick question...

For my Honours project I am running a risk factor analysis, therefore I've been using logistic regression. When using categorical predictors this whole process hasn't been a problem. However I have one variable that is 'Continuous' & this is giving me some grief in terms of how to explain/ report my output.

In terms of some background, the continuous predictor variable is Balance time (in seconds) & the dichotomous dependant is Injury (yes/no).

What I'm having trouble with is explaining/ interpreting the Odds Ratio in terms a unit change (which i want to be expressed as change in odds with every 1 second improvement or reduction in balance time).

The output in question is attached.

Any help/ explanation would be amazing.

Cheers

 

[lancearmstrong1313]

Are you modeling the odds of injury?

Assuming this is the case, you can interpret the odds of balance time as such:

For each second increase in balance time (holding any other variables constant), the odds of injury is .450 times as high (or decrease by 55%). Alternatively, you could say that for each second DECREASE in balance time, the odds of injury increase by about 2.222 (2.222 = 1/.450 - just taking the inverse).

 

[Rekansha]

Ahhhhh I see... For some reason I didn't think that the unit change was in seconds & I would have to do some crazy conversion.

Thank you for the help.

 

[sabine_m]

Thanks for this explanation! I was getting very confused about the interpretation of an OR below 1, but I think you just made it a lot more clear to me!

One remaining question (probably a stupid one, but I'm still a bit confused...): could you then also say "for each second INCREASE in balance time, the odds of injury DECREASE by about 2.222", so just flip around the signs?

Thanks!