BIOL2022 - L01
The University of Sydney
By the end of this lecture, you should be able to:
HATPC, used in several data science units at the University of Sydney.
A logical framework for ecological research proposed by Underwood (1997).
1Underwood AJ (1997) Experiments in Ecology: Their Logical Design and Interpretation using Analysis of Variance. Cambridge University Press, Cambridge.
A model-centred approach to experimental design from Fox et al. (2015).
3Fox, G. A., S. Negrete-Yankelevich, and V. J. Sosa. (2015). Ecological statistics: contemporary theory and application. Oxford University Press, USA.
There is no magical statistical method that will make up for a poorly designed study.
“A single poor design choice can make an experiment nearly worthless.”
“Far better an approximate answer to the right question … than an exact answer to the wrong question.”
We need to think about:
The relationship we expect is our starting model.
We may revisit these decisions as the study develops.
What does this represent? Licensed from Adobe.
This model retains enough information to describe the terrain without showing any value.
Models:
| Biological question | Predictor | Technique |
|---|---|---|
| Do two groups differ? | Two categories | t-test |
| Do several groups differ? | Several categories | ANOVA |
| Is there a “relationship”? | Continuous measurement | Linear regression |
Many of us have learnt statistics as a set of techniques, often relying some kind of decision tree.


The statistical model will eventually be a general linear model (GLM) with special cases, but we do not necessarily need to identify the “formal” name.
Heliocidaris erythrogramma lives on rocky reefs along the NSW coast.
Experiments show that its metabolic rate can be affected by:
For now, we will isolate one relationship:
How does water temperature affect metabolic rate?

We could expose urchins to a range of temperatures and ask how their metabolic rate changes.
Or we could choose a few temperatures and compare metabolic rate among the groups.
Both studies begin with the same biological idea, but they treat temperature differently.
The analyses will look different, but they share the same modelling framework.
| Variable | Role in the model | How we can represent it |
|---|---|---|
| Metabolic rate | Response | Continuous |
| Temperature | Predictor | Continuous or categorical |
\[\text{Metabolic rate} \sim \text{Temperature}\]
The design must give us data that can answer the biological question.
Begin with the relationship in our question:
\[\text{Metabolic rate} \sim \text{Temperature}\]
Account for another important biological variable:
\[\text{Metabolic rate} \sim \text{Temperature} + \text{Body size}\]
Add pH if it is part of the question or design:
\[\text{Metabolic rate} \sim \text{Temperature} + \text{Body size} + \text{pH}\]
We add variables because the biology and study design require them, not simply because they are available.
Did we do any statistics today?
We will cover more on models and study design in the next few weeks, but I hope you are less intimidated by the process!
Do not forget…
This presentation is based on the SOLES Quarto reveal.js template and is licensed under a Creative Commons Attribution 4.0 International License.