Fixed Coefficients (generalized linear mixed models)

Figure 1. Fixed coefficients for logit "No service"

In the viewer for the generalized logit mixed model, click the Fixed Coefficients view thumbnail.
From the Style dropdown of the Coefficients view, select Table.
Confirm that No service is selected in the Multinomial dropdown.
Click Exponential. This displays the exponentiated coefficients and confidence intervals for the exponentiated coefficients.
In the parameter estimates table, click the Coefficient cell. This displays the standard error, t statistic, and confidence interval.
The No service logit provides information about the probability of a customer answering "no service" as their usage relative to "service with company". For example, the parameter edcat=1 tells us that the odds of choosing “no service” instead of “service with company” for people in the low education category (at least high school diploma) are estimated to be exp(0.349) = 1.417 times the corresponding odds for people in the high education category (post-undergraduate degree), all other things being equal. This means that people with low education are more likely than people with high education to prefer “no service” to “service with company”.

Similarly, positive estimates for inccat=1 and inccat=2 imply that people in the low (under $45k) and median (between $45k and $100k) income categories are more likely to prefer “no service” to “service with company”. The negative estimate for Nop implies that the more people in the house, the more likely the person will prefer “service with company” to “no service”. This make senses because there is a cost advantage, per person, to having service when the number of people in the house increases.

Figure 2. Fixed coefficients for logit "Other provider"
Select Other provider from the Multinomial dropdown.

The Other provider logit provides information about the probability of a customer answering "other provider" as their usage relative to "service with company". The positive and significant estimates for edcat=1 and edcat=2 imply that people with low or median education are more likely to have service with other providers instead of having service with company.

In addition, the negative estimates for service_type=1 and service_type=2 in both panels mean that people are more likely to have TV or phone service with the company than internet service.