What Is a Reproducible Project?

Open workflows for ecology and wildlife research

Topics:

  • Folder structure
  • Raw vs processed data
  • Scripts and automation
  • Documentation
  • Portable workflows

Learning Goals

By the end of this session you should be able to:

  • Recognise the components of a reproducible project
  • Understand why project structure matters
  • Separate raw and processed data safely
  • Organise scripts logically
  • Use relative paths
  • Create a simple reproducible workflow in RStudio

Does this look familiar?

Desktop/
  analysis/
    final_analysis.R
    final_analysis_v2.R
    final_analysis_v2_FINAL.R
    final_analysis_v2_FINAL_REAL.R
    figure_new_final.png
    temp_data.csv
    old_data.csv

Common problems

  • Which script is correct?
  • Which data were used?
  • Which figures belong to which analysis?
  • Can another person rerun this?

Reproducibility Is Not Just About Publishing Code

Reproducibility means:

Someone else can:

  1. Open the project
  2. Understand the workflow
  3. Run the analysis
  4. Recreate the outputs

Ideally with minimal confusion

This includes:

  • Data organisation
  • Documentation
  • Scripts
  • Dependencies
  • File structure

Why Reproducibility Matters in Ecology

Ecology workflows are often complex

Examples:

  • Camera trap datasets
  • Spatial layers
  • Occupancy models
  • Meta-analysis workflows
  • Long-term monitoring
  • Multi-country collaborations

Complexity increases risk

Small workflow issues can make projects difficult to reproduce.

A typical reproducible workflow

Raw data
   ↓
Cleaning scripts
   ↓
Processed data
   ↓
Analysis scripts
   ↓
Figures and tables
   ↓
Reports/manuscripts

Core components of a reproducible project

A project usually contains:

  • Raw data
  • Processed data
  • Scripts
  • Outputs
  • Documentation
  • Metadata
  • A project file

The structure itself is part of the methodology.

Example project structure

my_project/
├── data_raw/
├── data_processed/
├── scripts/
├── outputs/
├── docs/
├── README.md
└── my_project.Rproj

Benefits

  • Easier navigation
  • Easier collaboration (even with your future self!)
  • Easier automation
  • Easier handover

Raw vs Processed Data

Never manually edit raw data

Raw data

  • Original source files
  • Untouched
  • Read-only if possible

Processed data

  • Cleaned data
  • Derived variables
  • Analysis-ready datasets

Example ecology workflow

camera_trap_data.csv
        ↓
clean_camera_data.R
        ↓
camera_data_clean.csv
        ↓
occupancy_model.R
        ↓
occupancy_results.csv
        ↓
figures/

The scripts explain exactly what happened.

Scripts Are Better Than Manual Steps

Manual workflows are fragile

Problems with manual editing:

  • Difficult to track changes
  • Easy to introduce errors
  • Impossible to reproduce precisely
  • Hard to scale

Scripts are documentation

Good scripts explain:

  • What was done
  • Why it was done
  • In what order

Documentation Matters

Good projects explain themselves

Minimum documentation:

A README.md should explain:

  • What the project is
  • Where the data came from
  • How to run the analysis
  • Folder structure
  • Required software/packages

Future-you will appreciate this.

Relative Paths

Avoid absolute paths!

Bad

read.csv("C:/Users/matt/Desktop/project/data.csv")

Better

read.csv("data_raw/data.csv")

Even better

here::here("data_raw", "data.csv")

Why Relative Paths Matter

Absolute paths break collaboration

Absolute paths:

  • Only work on your computer
  • Break when folders move
  • Cause problems across operating systems

Relative paths:

  • Improve portability
  • Improve collaboration
  • Improve reproducibility

By biggest bugbear

rm(list=ls())

This is just rude

Gold-standard reproduciblity

Red Fox IPM

Common Mistakes

  • Files on Desktop
  • Manual spreadsheet editing
  • Missing scripts
  • Missing metadata
  • Unclear filenames
  • Mixing raw and processed data
  • No version control
  • Hard-coded paths

Hands-on exercise

Goal

Build your first reproducible research project.

You will:

  1. Create an RStudio project
  2. Build a folder structure
  3. Add a small dataset
  4. Create a cleaning script
  5. Generate a figure
  6. Write a README

Exercise workflow

Create project
      ↓
Add folders
      ↓
Import raw data
      ↓
Create cleaning script
      ↓
Save processed data
      ↓
Create plot
      ↓
Document project

Step 1 — Create a project

In RStudio:

  1. File → New Project
  2. New Directory
  3. New Project
  4. Name it:
wildlife_reproducibility_project

Step 2 — Create folders

Create these folders:

/data_raw
/data_processed
/scripts
/outputs
/docs

Or create them using R:

folders <- c(
  "data_raw",
  "data_processed",
  "scripts",
  "outputs",
  "docs"
)

sapply(folders, dir.create)

Step 3 — Add raw data

Download the workshop dataset from GitHub.

Save it into:

/data_raw

Step 4 — Create a cleaning script

Create:

scripts/01_clean_data.R

Example:

library(readr)
library(dplyr)
wildlife <- read_csv(
  here::here("data_raw", "wildlife_data.csv")
)

wildlife_clean <- wildlife %>%
  filter(!is.na(count))
write_csv(
  wildlife_clean,
  here::here("data_processed", "wildlife_clean.csv")
)

Step 5 — Create a figure

Create:

scripts/02_make_plot.R

Example:

library(ggplot2)
library(readr)

wildlife_clean <- read_csv(
  here::here("data_processed", "wildlife_clean.csv")
)

p <- ggplot(wildlife_clean,
            aes(x = habitat,
                y = count)) +
  geom_boxplot()

ggsave(
  filename = here::here("outputs", "habitat_plot.png"),
  plot = p,
  width = 6,
  height = 4
)

Step 6 — Create a README

Create:

README.md

Include:

  • Project title
  • Short description
  • Folder structure
  • How to run the analysis
  • Required packages

Keep it simple

What did we build?

You now have:

  • A reproducible folder structure
  • Raw and processed data separation
  • Documented scripts
  • Reproducible outputs
  • Basic documentation