Sorry everyone – I know I said I’d start with the Completely Randomized Design BUT… the concept of the experimental unit keeps cropping up in some discussions. Turns out that as much as I would like to keep this straight forward for everyone, even statisticians have different views on this – oish!

So let’s take that step back and define an experimental unit – the unit to which a treatment is applied. Easy right??? Come on – straightforward? So let’s try out a few scenarios to see how straightforward it really is or is not… Fill out the table – and if you’re curious about the correct answers – email me your table and I’ll let you know how you did 🙂

So… there are a couple of things happening here – it is really easy for me to say what the experimental unit is – based on the unit to which we apply the treatment to. But, folks get confused when you start looking at the unit you take measurements on. In some cases I listed above there are several measurements taken on one experimental unit. For example – the second line in the table above – the experimental unit is the pot, since I am applying the fertilizer to the pot. However, I am measuring the 3 plants in the pot – these measurements are referred to as sampling units and NOT experimental units. We are always looking for ways to increase the sample size of a trial – but this is NOT the way to do it. I’ll talk more about the whys when we start looking at RCBDs.

The other challenge that I’ve encountered lately is defining the experimental unit in a way where we are trying to account for environmental variation by incorporating it into the experimental unit. We encountered this in a poultry trial – of course, I’ve been on vacation for a week now – and cannot remember the specifics – but it was related to aviaries in different rooms for behavioural and production trials. Rather than trying to remember – I’ll update later when I have the details. But a few of us discussed this and thought it was odd – even though the study was vetted and supported by a statistician, we did not agree with how the experimental unit was defined. Would the results have changed had the experimental unit been redefined? I don’t know – I suspect not! BUT, if you are going to do this, as a student PLEASE – be prepared to defend these choices! There is NOTHING worse than defending these choices by saying “The statistician told me to do this!” I cringe when I hear that some of my students have said this. This is your analysis and your work – ensure you understand what decisions were made. Be able to defend what your experimental unit was and how it was used in your analysis.

Any experimental design is dependent on the correct identification of the experimental unit. Work with a statistician to ensure you have a robust sample size to answer your research question. Recognize that there is no one RIGHT answer – be able to defend your choices!