How to best analyse ordinal data

Overcoming the simplification strategy

Analysing binary or continuous data usually doesn’t cause any headaches for statisticians. But when we step into ordinal data, most of us ignore their specific nature and either dichotomize them or analyse them as if they are continuous.

Recently, these problems have becoming much more prevalent due to the nature of composite endpoints (watch out for an interesting episode on this in a few weeks).

Now Benjamin and I have worked on better tools to analyse such data already at university. We’ll dig back into what we learned then and what is still relevant today.

Specifically, we’ll cover the following questions:

What are ordinal data?
How do we commonly analyse ordinal data and what’s wrong about these approaches?
What are better ways to analyse ordinal data?
How can we present ordinal data best using graphics?