Getting values larger than 1 for a logit test?

#1
I have run a logit test on a binary dependent variable "gambled?" against "Age" in STATA. I have the dependent variable coded as 0 for no and 1 for yes. However I have a coefficient of 1.54 on the independent variable "Age". How can this be in a logistic regression that is meant to be between 0 and 1.

Secondly, how can I interpret age here?

:)
 

noetsi

Fortran must die
#2
When you say the coefficient, do you mean the slope? If so that is not restricted to be between 0 and 1 or any value. It simply shows the change in the logit for a one unit change in the independent variable. It is only the dependent variable that has to have only two values (and they don't have to be 0 and 1, that is just common to code them that way. Either way they are not a range between 0 and 1 - they are just two specific values you coded).
 
#3
When you say the coefficient, do you mean the slope? If so that is not restricted to be between 0 and 1 or any value. It simply shows the change in the logit for a one unit change in the independent variable. It is only the dependent variable that has to have only two values (and they don't have to be 0 and 1, that is just common to code them that way. Either way they are not a range between 0 and 1 - they are just two specific values you coded).
Yes, on STATA it has the "Coef." which is slope coefficient.

Do you know how in this case you could interpret a continuous variable such as age against a binary dependent variable, with a value of 1.54?
 

noetsi

Fortran must die
#4
I would not use the slope in logistic regression since it has no intuitive meaning (it deals with the logit and few connect this to the underlying data). I would use the odds ratio which is far easier to intepret. It shows how the odds of being in the maximized state of the DV (which you choose) change for a one unit change in the IV.

This is one source for this .

http://www.ats.ucla.edu/stat/mult_pkg/faq/general/odds_ratio.htm

Slope is not the same thing as an odds ratio as hlsmith notes below. You can either calculate an odds ratio from the slope as he does, or you can look for it in the ouput.

For an interval variable there are several ways to interpret an odds ratio. Say the odds ratio was 2 for age. Than you could say for for a one year increase in age the odds of being in the level of the DV you are maximising is 2 times greater or equivalently that the odds went up 200 percent.
 
Last edited:

hlsmith

Less is more. Stay pure. Stay poor.
#5
I do not use STATA that often and wonder if the coefficient needs to be exponentiated to have the definition presented in post #4 (i.e., e^1.54)?