Logistic regression analysis was used to explain the relationship between response variables in the form of dichotomous/binary data with independent variables in the form of interval and or categorical scale data (Hosmer and Lemeshow, 1989).

Dichotomous/binary variables are variables that only have two categories, namely the category that states the event of success (Y=1) and the category that states the event of failure (Y=0). The Y response variable is assumed to follow the Bernoulli distribution.

The steps in using Logistics Regression analysis are:

1. Model Significance Test

To determine the effect of the independent variables on the dependent variables together (overall) in the model, the Likelihood Ratio Test can be used.

The hypothesis is as follows:

Ho: β1 = β2 =....= βp = 0 (there is no simultaneous effect of independent variables on dependent variables)

Hi: there is at least one βj ≠ 0 (there is an effect of at least one independent variable on the dependent variable)

2. Test Model Parameters

In general, the purpose of statistical analysis is to find a suitable model and a strong relationship between the model and existing data. Testing the significance of the parameter (coefficient ) partially can be done through the Wald Test.

The hypothesis is as follows:

Ho: βj = 0 (the independent variable j has no significant effect on the dependent variable)

Hi: βj ≠ 0 (the independent variable j has a significant effect on the dependent variable)

3. Odds Ratio

Odds ratio is a measure of risk or tendency to experience certain events between one category and another, defined as the ratio of the odds for xj = 1 to xj = 0. This odds ratio states the risk or tendency of the influence of observations with xj = 1 how many times if compared with observations with xj = 0. For independent variables with continuous scale, the interpretation of the coefficient j in the logistic regression model is that every increase in c units in the independent variable will cause the risk of Y = 1, which is exp(c.βj) times greater.