**When do we use a logistic regression?**

When we want to produce odds ratios to see if our independent variables (e.g. smoking: never smoker, ex-smoker, current smoker) predicts higher odds of the dependent variable (e.g. depression: yes or no). The outcome variable must have 2 categories.

**Example Scenario**

Computing the odds ratio of having depression based on people's smoking behavior.

In this scenario, our dependent variable is depression, and it has 2 categories:

1=No (reference category)

2=Yes

Our independent variable is smoking behavior, and it has 3 categories:

1=Never smoked (reference category)

2=Ex-smoker

3=Current smoker

Our research question is:

Compared to those who never smoked, do those who are ex-smokers and/or those who are current smokers have higher odds of having depression?

**Step 1**

Analyze-> Regression-> Binary Logistic

**Step 2**

Select the dependent variable (depression) and move it into the Dependent box. Move the independent variable (smoke3) into the Covariates box.

**Step 3**

Click the Categorical... box. Move smoke3 into the categorical covariates box because smoke3 is a categorical variable (do not need this step if your independent variable is a continuous variable). Select First as the Reference Category and click change, because we want the first group (never smoked) to be the reference category.

**Step 4**

Click the Options... box. Tick CI for exp(B): 95% -> this will give you 95% confidence intervals for your odds ratios

As the 95%CI do not overlap, we can conclude that compared to those who have never smoked, ex-smokers have 1.14 times higher odds (95%CI=1.05 to 1.24), and current smokers have 1.79 times higher odds (95%CI=1.64 to 1.95) to be depressed.

© Maths-Statistics-Tutor.com 2010 Web Development Team.