I want to look at the relationship between 1 response (1-5 likert scale) and a dichotomous outcome variable (here/not here). What is the best way to do that?
2 Answers
$\begingroup$
$\endgroup$
research logit and probit. They fall into the category of "limited dependent" approaches, and are often used when you have an independent variable and want to develop a statistically based model that gives you a yes/no outcome
$\begingroup$
$\endgroup$
You may use the Kruskal-Wallis Rank Sum Test. SO that you can check whether the dichotomous outcome variable is associate with the response. As the Kruskal-Wallis Rank Sum Test does not assume normal distribution, it fit the likert scale quite well. Nonetheless, if you strongly believe the scale is evenly spaced and normally distributed, a simple t-test can be used.
