Logistic regression model - numeric predictors modification

[Martijncats]

I am running a regression using multiple variables in order to predict a binomial outcome. Two of my variables are numeric, most are categoric. My two numeric variables are age and bmi. I have calculated an odds ratio for each of those variables (e.g. if the bmi increases with 1 point, possibility of dependent variable changes). My goal is now to calculate an odds ratio and Low/High 95% CI for those variables, if the bmi and age increase with 5 points/years instead of 1. How do i calculate this?

 

[hlsmith]

You can divide all numeric values by 5 and use that variable in the model. So then when you think about the 1 unit increase it will actually be 5 units. Let me know if this doesn't answer your question.

 

[fed2]

I think you can also take the odds ratio to the 5th power, the R package 'library("oddsratio") does this automatically.

 

[hlsmith]

I have never heard of this action, I will have to try it. I know you can flip the outcome by inverting the estimate and there is a wonky conversion between ORs and RR using sqrt given the outcome is rare or not.

 

[fed2]

I would not tell a lie. Its apparently not a widely appreciated thing.


Edit: I might lie, by ommission or carelessness. remember the opposite of what i said is equally possible.

 

[hlsmith]

Boom. I am a person of doubt at times. So I just went out of my way to tested this and it holds. Of note, one could also use the variable's standard deviation as the value.

 

[Martijncats]

Thank you a lot for your help guys! Both ways work perfectly! Cheers