Comments on reproducibility of the code

[chronchi]

Hi, I'm Carlos from the workshop on open science at EPFL/ETHZ (June 2022). First congrats on the paper, it is always a lot of work and it is very nice you shared the code. Also it is very nice that you provided a license.

I have some comments and feedback to you here. I separate the comments in 3 sections. First section some issues regarding the scripts, second about R packages and third some suggestions on what you could do here, from my point of view.

Running the scripts

It is also not possible to run the scripts. There are some reasons for why I could not run it:

• There are two folders without any description of what they are, probably someone with domain knowledge know it
• Files to run the analysis are not available in the repo and not indication where they can be downloaded. It seems that in one of the scripts you are using Rdata, no mention from where it comes. To be more specific it is the following line:

  Plastic_PMFA-Emission/Emission/5-CalculationScript.R

  Line 18 in bc5773f

  load("Input/InputReady.Mod2.Rdata")

• Also in the same script, the folder "Code" is being referenced:

  Plastic_PMFA-Emission/Emission/5-CalculationScript.R

  Line 10 in bc5773f

  source("Code/functions.needed.R")

  
  so it might be better to just change the Code to the current folder, as I see that the script is there.
• File paths are all absolute paths
• No mention about the libraries being used
• No mention of which R version is being used
• No header in the scripts, what they do

R package - documentation

You mention this is a R package for the paper. By looking at the scripts and folder structure, this is not a R package. The structure for a R package is missing. A good website that says how to produce a R package is https://r-pkgs.org/.

Suggestions

I think your project would benefit from a folder structure for data science projects. For this you don't need to make a R package. My recommendations are the follow:

• Create three top level folders: data, scripts and results
• In the scripts folder, add some subfolders and organize the scripts as you seem fit
• Put all the raw data in the data folder
• The results folder is where you save the analysis results, such as images and even intermediate rds files
• Do never use Rdata to save results, use only rds files to save intermediate states, this way you are explicit on what you are saving. If you are using Rdata, you are saving the environment and this way it is impossible to know what you are loading by just looking at the script. Also this increases the chance of having bugs and unexpected results
• When running a R project, start a project directly from Rstudio. This makes it easier to work across different projects. Here is a link with more information for this: https://support.rstudio.com/hc/en-us/articles/200526207-Using-RStudio-Projects
• If your project is self contained, try to use relative paths, this way any person could download the files and run the scripts. Provided they keep the folder structure.
• In each script, try to include headers with the following information: author, email, date (when you did the analysis) and a brief description of what the script is doing.
• Another option when using R is to use quarto (https://quarto.org/). This way you can mingle R with markdown, getting very beautiful reports. This way you can mix code and text, where you can explain more what is happening and why you are doing this. I have a repo here with an example: https://chronchi.github.io/transcriptomics
• When documenting the functions, try to follow the Roxygen guidelines, check the r-pkgs website, they explain it more there (https://r-pkgs.org/man.html). What is very nice from doing this is that you get the vignettes in either pdf or html automatically. No need to write more stuff besides what is in the script.
• For the libraries you can have a requirements.txt file stating which libraries you are using

Others

If you have any question or want to discuss more, feel free to drop me an email as well at carlos.ronchi@epfl.ch, besides replying to this issue.

I wish you good luck with your research!

[L-Zipeng]

Hi Carlos, thank you for providing such detailed and professional advice and reviews. Here are the replies corresponding to your suggestions.

Background

This coding package is the first project I used R and the first file I uploaded on GitHub in my master’s study. At that time, I have no idea how I should organize the coding structure and how to upload the scripts in the correct format. For sure, there are many mistakes to correct and places that need to be improved. Thank the EPFL open science workshop for giving me a chance to review my code and receive suggestions from you.

Scripts

The scripts only serve as supplementary codes processing the modelling data from the paper. Therefore, they are a bit hard to understand directly. I have uploaded the input data file as well as the example results.
The PMFA folder is the first part of the model, calculating the plastics' probabilistic material flow analysis results based on their detailed lifecycle with the input data, including production, trade volumes and transfer coefficients of each flow pathway. The Rdata files are generated by each step of the data processing steps, which should be available with the input data file.
The Emission folder is the second part, which uses the Rdata files from PMFA folder, generating the emission calculation based on PMFA.
The formatting issues about absolute paths hears and explanations are problems that I will pay attention to in the future.

Documentation and suggestions

Thank you so much for pointing out the definition of the R package and following suggestions. I learned R by myself and did not have systematic knowledge, the suggestions are super helpful, and I will follow them in later projects. I learned a lot from your detailed suggestions. Thank you very much!