Solving the Midpoint Melee: Introducing New Methods for Plant Cover Classes

Post provided by KATHI IRVINE and TOM RODHOUSE

Collecting ordinal data. ©NPS

Or better yet, this post could be named ‘Our Cathartic Journey to Convince Ecologists to STOP Using the Midpoint Values for Analysing Plant Cover Classes’. Our work picks up where another recent post (Stuck between Zero and One) and Methods in Ecology and Evolution article (‘Analysing continuous proportions in ecology and evolution’) by Douma and Weedon left off. They introduced the benefits of using beta and Dirichlet regression. We’re going to tackle the sticky wicket of ordinal data. So, what should you do if you assign a range (like 0.2 to 0.3) instead of record a value (like 0.22) for a continuous proportion?

What is Ordinal Data?

It’s probably a good idea to start by defining the type of data we’re talking about. The best example is from plant surveys. Biologists visually assess the percentage of a pre-defined area covered by a certain plant species. They then record a ‘cover class value’ as an estimate of abundance. Each cover class value corresponds to the percentage of the area that is taken up by the plant in question (e.g., record a 0 for 0%, record a 1 for >0-5%, record a 2 for >5-25%, …, record a 6 for >95%). Continue reading