Lecture 01 — Introduction

Our question, our data, and a first look at what we will do

Bill Perry

2026-07-05

Lecture 1 · Syllabus

The syllabus has all the details — topics, grading, policies, schedule
Read it before Thursday
It will evolve — check Canvas for the current version

Note

✅ Key idea

The syllabus is the contract between us. If something is unclear, ask now — not at exam time.

Lecture 1 · Who am I?

Bill Perry — call me Bill or Dr. Perry
Office: Swenson Science Building 13
Phone: 218-726-8145
Email: wlperry@d.umn.edu
Drop-in hours: Monday 12:15–1:15, Tuesday 2–3 (tentative)
Lab: SSB 170

Tip

Best way to reach me: email or drop in. If I am in the lab I am usually around.

Lecture 1 · My goals for this course

Help you learn the tools to organize and present data from potentially very large datasets
How to use Excel for some tasks — and when to stop using it
How to use R — a coding language — to work with data and graphics
How to organize, clean, summarize, and visualize data — the pipeline
How to formulate and test hypotheses
How to use statistics to test those hypotheses
How to interpret what the data is telling us

Note

The pipeline: raw data → clean → summarize → visualize → test → report This course teaches every step.

Lecture 1 · My expectations

Communication — ask questions in class, in office hours, by email
Practice — the only way to learn to code is to write code
Failure — every error message is information; celebrate finding bugs
Learn to troubleshoot — reading error messages is a skill we will practice

Important

Getting stuck is not failing — it is the job.

Every working data scientist searches Google for error messages every single day.

What I will do:

Provide clear examples
Walk through code line by line
Give you real datasets to work with

What you need to do:

Type the code yourself (do not copy-paste)
Run it, break it, fix it
Bring questions to class

Lecture 1 · Science

Science is a way to acquire, organize, and apply knowledge
We make predictions and test them with a falsifiable approach — statistics
Explanations that cannot be tested or falsified are not science

Example:

“Leaves on the shady side are bigger” testable
“The forest has a spirit that controls leaf size” — not testable

Note

📖 New vocabulary

Falsifiable = a prediction that could, in principle, be shown to be wrong.

A good hypothesis is one you could prove false.

Today · Real-world data science

Talking about data science is super boring
The only way to learn it is to do it — collect data, enter it, analyze it
This lecture and the next will be about how to work with data from the very start
Your task: collect data → organize it → enter it into a spreadsheet → develop metadata

Tip

🖐 Try it yourself

We are going to go outside, collect leaves, and build our first dataset today. That dataset will follow us for the first part of the course.

What will we do

We are going to measure something in nature
Suggestion: leaf mass, width, length, and surface area from the sunny side vs. shady side of a tree
- Do leaves respond to light intensity?
The null hypothesis (H₀):
- There is no difference in leaf size or shape between the two sides of a tree.
The alternate hypothesis (Hₐ):
- Leaf size differs between the sunny and shady sides.

Note

📖 New vocabulary

Null hypothesis (H₀) — the “nothing is happening” claim. Statistics will tell us whether the data are consistent with this or not.

Alternate hypothesis (Hₐ) — the claim we think might be true.

How does leaf size vary?

Does the sunny side vs. shady side of a tree differ due to light?
- Predict: what do you think will happen before we collect any data?
  - Your prediction:
- Hypothesize: what is the formal testable statement?
  - Your null hypothesis: ____________________________
  - Your alternate hypothesis: ____________________________

Tip

Key distinction

A prediction says what you think will happen and usually has a direction.
A hypothesis is a formal statement that can be accepted or rejected with data.

How would we collect data to test this?

Can we collect leaves from one tree to do this?
- YES? or no NO?
- Discuss
What is the problem with using only one tree?
How might results from many trees differ from one tree?
Do we need to be sure our sample is randomly selected?
- Why do we need to randomize?
- How can we randomize our sampling?

Important

Replication and randomization

These two principles are the foundation of all experimental design. One tree is a sample size of one — we cannot generalize from that.

What variables could we record?

What variables could we record about the tree?
What variables could we record about the leaf?

Dependent variable (response): - The thing we measure — it depends on our treatment - Example: leaf mass, leaf area

Independent variable (explanatory): - The thing we control or group by - Example: side of tree (sunny vs. shady)

Note

📖 New vocabulary

Dependent variable → what you measure (Y)
Independent variable → what you control or group by (X)
- In a graph: Y goes on the vertical axis, X goes on the horizontal axis.

Each group needs to collect at least 2 pairs of leaves

Collect 1 mature leaf from the sunny side and 1 from the shady side of the same tree
Do this on at least 2 separate trees (we share trees since we are limited)
Standardize your collection: (DISCUSS AND FIGURE IT OUT BY YOURSELVES)
- How high do we collect from?
- What does “mature” mean?
- Any other decisions to standardize?

Important

⚠️ Watch out!

If everyone collects differently (different heights, different leaf ages), we cannot combine our data.

Agree on the rules or methods before collecting…..

Back in the lab · Before we begin

What can we measure?
- mass (grams)
- length (mm or cm)
- width (mm or cm)
- surface area (cm²) — how?
- others?
Decisions to make as a group:
- What units?
- What abbreviations or vocabulary for the variables?
- What if a leaf is damaged?

Note

📖 New vocabulary

Metadata = data about your data.
Who collected it? When? Where? With what instrument? Under what conditions?
Without metadata, data are nearly useless to anyone else — including future you.

Back in the lab · Setting up the spreadsheet

How do you enter this into the Excel sheet?
- What does the sheet look like? Mock something up
What are the column names you will use?
What are the units for each column?
What is the vocabulary? (no spaces in column names, use _ instead)
Where does the metadata go?

Tip

Best practice for column names:

leaf_mass_g, not Leaf Mass (g) - why???

Consistent, no spaces, no special characters, units in the name and all lower case - why?????

Can you make a graph in Excel?

Using what you collected, make a simple Excel chart
- What would the graph look like?
- What goes on the X axis? The Y axis?
- What can you derive from the data already?

Note

Excel is great for a first look. But as datasets grow larger and analyses more complex, we need another approach - R.

What R code looks like · A preview

You do not need to understand this yet — just look at the shape of it.

# Load the tools we need
library(tidyverse)
library(readxl)

# Read our leaf data into R
tree_df <- read_excel("data/tree_experiment.xlsx")

# Compute the mean weight for each side
tree_df %>%
  group_by(side) %>%
  summarize(mean_weight = mean(weight_g))

# Make a plot
ggplot(tree_df, aes(x = side, y = weight_g)) +
  geom_boxplot()

Tip

🖐 Notice

It reads almost like English: take tree_df, then group by side, then compute the mean. That is the goal — code that tells a clear story.

Before Thursday · Install R and Positron

Downloading R and Positron:

Download R — cran.r-project.org
- Installs to your computer — can you find it?
Download Positron — positron.posit.co
- Install R first, then Positron — Positron needs to find R on your machine

Important

⚠️ Watch out!

Install R first, then Positron. If you install Positron first, it may not find R automatically.

See you Thursday — bring your laptop and the leaf data you entered in Excel.