Regardless of whether you are performing a simple or multiple multinomial regression analysis, you always need to choose a base outcome for the y-variable. Our results will only show us to differences between the base outcome and the other categories of the outcome, and not contrast the remaining combinations. Of course, you can repeat your analysis and alternate between the base outcomes – or you can use the listcoef command. This command requires that you install a user-written package first. So, if you have not installed it already, type:

search spost13_ado

Click on the first link in the list, and then choose Click here to install.

Let us go back to the multiple regression analysis that we conducted earlier. There, we chose Married as the base outcome. The quietly option is included in the beginning of the command to suppress the output.

quietly mlogit marstat40 gpa sex ib1.educ if pop_multinom==1, rrr b(1)

We move on to the listcoef command (which is a postestimation command). Beware that this often produces comprehensive output.

listcoef

In the table above, you can see the differences between all base outcomes, stratified by the x-variables included in the model.

Note The coefficients (in the column called b) are relative log odds, not relative risk ratios.

It is possible to make the output a bit more compact, for example by specifying the pvalue option:

listcoef, pvalue(05)

With the above-specified option, only variables/categories with a p-value below 0.05 will be displayed in the table.

More information help listcoef

You can get some further assistance in interpreting the coefficients in the table by the mlogitplot command. This command requires that you install a user-written package first. So, if you have not installed it already, type:

search spost13_ado

Click on the first link in the list, and then choose Click here to install.

Let us go back to the multiple regression analysis that we conducted earlier. The quietly option is included in the beginning of the command to suppress the output.

quietly mlogit marstat40 gpa sex ib1.educ if pop_multinom==1, rrr b(1)

The mlogitplot command creates an odds ratios plot for mlogit. We should specify the amount of change when we order the mlogitplot. In our multiple regression model, we have three variables. The first is gpa: this is a continuous variable which can be described using unit change (amount option: one). The second is sex, which is a binary variable that can be described using binary change (amount option: bin). The third is educ, which is included as a factor variable – this automatically plots changes from 0 to 1.

mlogitplot gpa sex ib1.educ, amount(one bin)

The letters denote the different categories of marstat40. On the y-axis, we have the variables/categories included in the model. The x-axis shows the odds ratios (as well as the relative log odds) in relation to the base outcome (which is Married, or M, in this case).

Note Between some of the letters, there are lines. This means that the difference is not statistically significant.

More information help mlogitplot