Logistic regression coefficient help

[BenInUtah]

I have a logit that I have run. my independent variable can take whole number values from 0 to 5. I got a coefficient of ~ 0.48, which when I plug into the equation e^0.48, I get a value of ~1.61, which doesnt make any sense to me. If I increase from 1 to 2 in my independent variable, I am 61% more likely to see a 'success'? Same with 2 to 3? Am i understanding this correctly?

 

[hlsmith]

Yup, what do the confidence intervals look like? Do you think there isn't a linear effect? If so, perhaps explore a GAM regression.

 

[fed2]

sounds right. What are the percentages in each of the 5 categories? They should line up with he 1.61 figure, more or less. If not, may be poor model fit.

 

[BenInUtah]

@fed2 can you remind me how to calculate those?

 

[BenInUtah]

@fed2 nevermind, i think I figured it out. I have an endogeneity problem and so I think i am going to use an instrumental variable approach.

 

[hlsmith]

Finding a true instrumental variable can be trick. So make sure you meet the assumptions.

 

[BenInUtah]

Finding a true instrumental variable can be trick. So make sure you meet the assumptions.

Found that as well and so I scrapped it.

 

[noetsi]

I have never been tempted to do instrumental variables.

What do you mean exactly by " I have an endogeneity problem"

 

[hlsmith]

It is econ speak for confounding.

 

[BenInUtah]

It is econ speak for confounding.