exercise2.md

Exercise: managing a Julia research project

In this exercise, we'll explore how the relationship between the size of a tree and the amount of carbon it stores changes from one species to the other. We'll use a simple linear regression model to estimate the slope of the relationship, and compare it throughout the species investigated.

Requirements

Part 1: Package management

• Activate your environment, and install the following dependencies
  • GLM
  • DataFrames
  • CSV

Part 2: project management

• Create a new project directory and set up the following directory structure

my_project/
├── src/
│   └── regression_functions.jl
├── tests/
│   └── test_regression_functions.jl
├── data/
│   └── species1.csv
│   └── species2.csv
├── results/
├── Manifest.toml
├── Project.toml
├── .gitignore
├── run_code.sh
└── my-analysis.jl

• In the src/regression_functions.jl, create a function linear_regression that takes in a string corresponding to the location of a CSV file, and outputs the slope associated p-value of the linear regression between tree_size and carbon_content variables in the CSV.

Hints

The slope of a GLM model may be retrieved as

slope = GLM.coeftable(model).cols[1][2]

And its p-value as

pval = GLM.coeftable(model).cols[4][2]

• Create a unit test for linear_regression in test/regression_functions.jl

  • Use the Test module

• In my-analysis.jl file, write a Julia script that

  • loads the necessary packages
  • loads the functions in src/regression_functions.jl
  • creates an empty DataFrame

    • df_results = DataFrame(species_name = [], slope = [], pval = [])

  • loops through the CSV files, runs for each a linear regression using linear_regression, and pushes the results to the DataFrame
    • Make sure to print some logging information, e.g. println("processing ", csv_filename)
  • exports the dataframe as a CSV file in the result folder

• Use a shell script to run my-analysis.jl

Solutions

You may find the solutions in the Day1/23_practical-guidelines/my_project_solutions folder.