Logistic regression is a type of statistical analysis used to find associations between a dichotomous outcome variable and other variables. Although there can be only one dependent variable in a logistic regression, the number of independent variables is not restricted to only one. Thus, using logistic regression, we can find associations between any two variables, while adjusting for others. There is also no restriction of dichotomy among the independent variables.

Logistic regression produces estimates and p-values corresponding to the association between the dependent variable and each independent variable. It also produces the odds ratios for each independent variable as compared to the dependent variable.

Usually logistic regression is used when the dependent variable (also known as the outcome variable) is binary. In some cases the outcome variable is not binary but ordinal. Recall that an ordinal variable has multiple levels (not just two) which follow a certain order. For example, a variable age group may have levels: young, middle age, and old. 

Ordinal Logistic Regression uses proportional odds to model ordinal data. The basic idea of this model is to recalculate the ordinal variable into a number of binary variables. Thus, instead of having one variable with three levels (young, middle age, and old), we model three binary variables, one for each level of the original variable. 

The interpretation of the ordinal logistic regression is not much different from binary logistic regression. It is important to note that valid interpretation of the analysis depends on whether proportional odds assumptions have been verified.

Contact Statistical Consulting Network for help with your logistic regression analysis