The obvious advantage is that L is a linear (additive) function.

Reversing the logit transformation allows us to examine the probability p that the event did occur:

p = 1 / (1+e-L).

The following series of tables presents the results from the logistic regression analysis on Internet access in the past three months and Internet access at home, with the following key demographic characteristics as dependant variables: education, age, income, gender and location. Gender and location are both categorical variables (male/female and urban/rural).

These results have been transformed to reflect the exponential (e) base. If e^B is greater than one, then B is a positive value (as X increases, Y increases). A value of e^B that is between zero and one means B has a negative value (as X decreases, Y increases). If e^B = 1, B = 0 and there is no contribution from X on the value of Y.

Regression Models with Demographic Characteristics

In the first table (Table A1), Internet access in the past three months in the 1999 survey, shows a significant contribution from all five variables on the likelihood of Internet access. Income (most important factor overall), education, and an urban setting all have a positive effect. The generational effect is also quite significant, with younger respondents more likely to have had access to the Internet in the past three months. The gender gap also shows up as significant, with men having an increased likelihood of having had recent access to the Internet.