Welcome to your very first BEDA practical. Today we will help you get started by giving you a free, physical lab notebook and making sure you have all the software you need for this unit.
By the end of this session, you should be able to:
We will explain to you how BEDA practicals work, what you will be doing, and how to get the most out of them. Your lecturer (probably Januar) will cover:
Each practical in Module 1 begins with a quick workshop. We will cover a key idea, then you will get stuck into some quick hands-on exercises in the lab. These are your chance to practise what you have heard in lectures and try out study design and data analysis for yourself.
Today’s workshop covers:
The workshop should take about 45 minutes unless Januar gets carried away. Hopefully one of you will keep him on track, but if not, a demonstrator will be there to remind him to wrap it up.
This exercise introduces Cheatsheets: handy guides with common functions and commands to help you get started with new software.
This is a good opportunity to compare different software for the same task. For example, try using both R and Jamovi to plot the same data. You might discover a new preference.
Your feedback helps improve these resources. Once you have tried a cheatsheet, please use the quick feedback form on the cheatsheets page. It only takes a minute. You can also request new cheatsheets, which Januar will create as the unit progresses.
If you would like to contribute directly, we welcome your suggestions. You can:
If your contribution is accepted, you will be credited as a co-author. This is a great way to contribute to the unit’s resources and something you can mention on your CV. You can contribute at any time during the semester.
One of the most important parts of study design is being able to model data. One way to do this is with plots – they are not just pretty pictures, but actually capture the heart of your study design. In this exercise, you will get to grips with what a model is and how it helps you plan and analyse experiments.
A model is just a simple way to represent something complicated. In data analysis, a model helps us make sense of a dataset, test ideas (hypotheses), or make predictions. Models can be as simple or as fancy as you like, depending on your data and your research question.
Soon, you will be using what is called an empirical model. It might look like a scary equation, but you do not need to be a maths genius to get the idea.
Here is an example:
\[y = \beta_0 + \beta_1 x_1 + \epsilon\]
This just means: “the value of \(y\) is equal to a constant, plus \(x_1\) times another constant, plus a bit of random error.”
But we can make it even simpler by thinking of it as a relationship:
\[y \sim x_1\]
This means: “\(y\) is influenced by \(x_1\).”
So if \(y\) is weight and \(x_1\) is age, it is:
\[\text{weight} \sim \text{age}\]
In other words, “weight depends on age.”
For now, let us focus on graphical models – in other words, plots! Plots are models because they show relationships between variables. For example, you could plot the weight of an animal species against its age or height to see if there is a pattern:
set.seed(328)
Weight <- c(rnorm(50, mean = 50, sd = 3), rnorm(50, mean = 60, sd = 3))
Age <- c(rep("Young", 50), rep("Old", 50))
# create data frame
data <- data.frame(Weight, Age)
# ggplot
library(ggplot2)
ggplot(data, aes(x = Age, y = Weight)) +
geom_boxplot() +
theme_classic() +
scale_fill_brewer(palette = "Set1") +
scale_x_discrete(limits = c("Young", "Old"))
Interestingly, how we consider your variables can drastically change the type of plot and model used for data analysis. For example, consider the same relationship as above, but plotted differently:
V_max <- 70 # Assume the maximum weight is 70 kg
K_M <- 10 # Assume the Michaelis constant is 10 years
# Generate age data
age <- seq(1, 20, by = 0.5) # From 1 to 20 years, in 0.5 year increments
# Apply the Michaelis-Menten equation to simulate weight data
weight <- V_max * age / (K_M + age) + rnorm(length(age), mean = 0, sd = 5)
# Create a data frame
data <- data.frame(Age = age, Weight = weight)
# Plot using ggplot2
library(ggplot2)
ggplot(data, aes(x = Age, y = Weight)) +
geom_point() + # Use geom_line for continuous data
geom_smooth(formula = y ~ x, method = "lm", se = FALSE) +
theme_classic()
Notice how both plots show the same kind of relationship, but the way we interpret them – and the type of model we use – is different. This is what study design and data analysis are all about: the model you choose can totally change your results, so planning ahead is key!
Your goal is to practise thinking like a research biologist. For this, you will choose a dataset, use that variables to formulate a research question, and sketch a graphical model to help answer your question. Remember, you do not need to analyse or process any data for this task. Simply knowing the type of variables you are working with (e.g., categorical, continuous) is enough to decide on and sketch an appropriate plot.
Explore the data. Download penguins.csv and possums.xlsx. Open them and familiarise yourself with the variables available in each dataset.
Formulate a research question and choose a model. Pick one of the datasets and think of a simple research question you could ask. Then, decide which type of plot would be the best way to visualise the answer. Here are some common plots and what they show:
Sketch your model and write down your question. In your new lab notebook, write down your research question clearly. Beneath it, sketch the plot that represents your model. For example, if your question is “How does possum weight differ between sexes?”, a boxplot would be a suitable model.
Think about your model. Consider the variables you have chosen. Why is your chosen plot suitable? Could you use a different plot if you collected the data differently? Discuss your ideas with your peers or your demonstrator.
A model comparing the weights of juvenile and adult possums could be shown as a boxplot or barplot. Here’s a boxplot of the possum data:
library(readxl)
possums <- read_excel("assets/possums.xlsx", sheet = 2)
library(ggplot2)
ggplot(possums, aes(x = Age, y = BW)) +
geom_boxplot() +
theme_classic()
See how thinking about how to interpret the data is a big part of study design? It helps you decide what data to collect and how to analyse it. But notice – there is only one juvenile possum in the dataset, so you could not really analyse this properly!
Food for thought: Can age still be used to construct a model given that there is only one juvenile possum?
That is all for today! If you have any questions, just ask your demonstrators – they are here to help. Do not forget to give feedback on the cheatsheets and have a go at the exercises in your own time. See you next week for more on study design and data analysis!
This section is for those who finish early or would like extra practice at home. Use the penguins and possums datasets from this practical. Don’t forget that you are either drawing plots or writing
Penguins dataset: Create a plot to show the relationship between bill_length_mm, bill_depth_mm, and flipper_length_mm. Is it possible? What model are you constructing and can you put it into words?
Possums dataset: Investigate the relationship between body weight (BW) and the personality trait for boldness (BoldBLUP). How do Sex and Age influence this relationship? How would you model three variables together in a plot? Can you model all four of them (BW, BoldBLUP, Sex, and Age) together?