A framework for classifying and visualising experimental designs when subjects are measured repeatedly
This paper addresses confusion surrounding repeated measures in experimental research by defining key design characteristics based on experimental units, measurement frequency, and randomization strategies to establish a framework for classifying and visualizing such designs using Hasse diagrams.
Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are a detective trying to solve a mystery, but instead of looking for clues at a crime scene, you are looking for answers in the behavior of living things—people, dogs, or even tiny cells. In the world of science, researchers often run experiments to see how different things, like a new medicine or a specific training routine, affect these subjects. But here's the tricky part: sometimes, they don't just check the subject once and move on. They might check them five times a day, or give them three different treatments in a row, or even take three tiny samples from the same blood draw. This is what scientists call "repeated measures."
Think of it like tasting a soup. If you take one spoonful, that's a single measurement. But if you take a spoonful every ten minutes to see how the flavor changes as it simmers, or if you taste it with three different spoons to make sure you aren't imagining things, you are doing "repeated measures." The problem is that scientists have been using the same label, "repeated measures design," for all these different scenarios, even though they are actually very different kinds of experiments. It's like calling a bicycle, a skateboard, and a unicycle all "wheeled vehicles." They all have wheels, but you ride them differently, and if you try to ride a unicycle like a bike, you might fall off. In science, if you analyze your data the wrong way because you misunderstood the design, your conclusions could be completely wrong, leading to bad medical advice or wasted research.
This paper, written by Damianos Michaelides and Simon T. Bate, is like a new instruction manual that finally sorts out the confusion. The authors noticed that people were mixing up different types of experiments where subjects are measured more than once, and this was causing errors in how the data was analyzed. They propose a clear framework to classify these designs based on three simple questions: Who is the main character (the experimental unit)? Are we checking that character over and over again? And most importantly, how did we decide who gets which treatment (the randomization strategy)?
Instead of just listing rules, the authors introduce a visual tool called a "Hasse diagram." Imagine a family tree, but instead of showing who is related to whom, it shows how different parts of an experiment are connected and how the treatments were assigned. By drawing these diagrams, the authors show that what looks like a messy pile of repeated measurements can actually be sorted into six distinct categories: standard repeated measures, within-subject designs, block designs, split-plot designs, cross-over designs, and nested designs.
The paper argues that you cannot just slap a generic "repeated measures" label on any experiment where someone is measured twice. For example, if a dog gets a pill every day in a specific order that cannot be changed, that is a "within-subject" design where the order matters. But if a human gets a pill, and then the order of the pills is randomly shuffled for each person, that is a "block" design. The authors show that treating these two different scenarios as the same thing is a mistake. They demonstrate that by correctly identifying which of the six categories your experiment fits into, you can build the right mathematical model to analyze the data.
The paper doesn't claim to have discovered a new type of medicine or a new law of physics. Instead, it offers a better way to organize the tools scientists already use. It suggests that by using these specific characteristics and the visual Hasse diagrams, researchers can avoid the trap of using the wrong statistical methods. The authors show through several examples—from clinical trials with humans to studies with rabbits and dogs—that when you get the classification right, you get the right answer. They don't say this fixes every problem in science, but they do suggest that it clears up a lot of the confusion that leads to misleading results. In short, they are handing researchers a better map so they don't get lost in the forest of their own data.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.