From research software engineering to data science: how technology will impact the way we research
Slides
https://www.flor14.github.io/wsu-dandrea/slides.html
This work is licensed under a Creative Commons Attribution 4.0 International License.
Title slide illustration: The Turing Way Community, & Scriberia. (2020, March 3). Illustrations from the Turing Way book dashes. Zenodo. http://doi.org/10.5281/zenodo.3695300
Hello
Dr. Florencia D’Andrea
lab 
field work 
Software quality can affect research results
What are the potential consequences of developing bad quality software?
Reproducibility crisis
Code unavailable is one of the reasons why researchers can’t reproduce their articles
Reproducibility
A result is reproducible when the same analysis steps performed on the same dataset consistently produces the same answer
Reproducibility
- Empirical
- Statistical
- Computational
Computational reproducibility
When detailed information is provided about code, software, hardware and implementation details.
Example
Situation 1
Julia wants to share R code with her supervisor
Plot’s code
Julia is running this code on her computer ✅
Show the code
library(readr)
library(tidyr)
library(ggplot2)
coffee_data <- read_csv("data/coffee_data.csv")
coffee_data |>
pivot_longer(cols=aroma:moisture,
names_to="caracteristicas_cafe",
values_to="value") |>
ggplot(aes(value, total_cup_points)) +
geom_smooth(method= "gam", span=0.3, color='purple') +
facet_wrap(~caracteristicas_cafe, scale="free_y")
Computational environments
Important to capture them to avoid:
Problems when sharing your code
That you code breaks with time
The sleight-of-hand trick that can simplify scientific computing
Computational environments and the tools to manage them can help researchers to deliver code that is reproducible, documented and shareable.
Situation 1-bis
Julia wants to run the code 10 years later
Challenge to scientists: does your ten-year-old code still run?
‘Missing documentation and obsolete environments force participants in the Ten Years Reproducibility Challenge to get creative.’
Computational environments
Important to capture them to avoid:
Problems when sharing your code
That you code breaks with time
Tools for reproducibility
We can use different tools to ensure the reproducibility of a project such as:
Situation 2
Julia is writing code for an article. What is the best way to share the code with other colleagues?
Git
A version control system 
GitHub
Repository hosting service 
Why Git + GitHub is the best way to share code?
It’s possible to assign names to each code version and revisit specific points in its history.
Your code is backed up online on GitHub.
It is easy to track the authorship of each change made
Why Git + GitHub is the best way to share code?
You can share your code as an open source project
Foster collaboration
- pull requests
- project boards
- github issues
Many research groups have their GitHub organizations online
:::footer https://github.com/dib-lab :::
Situation 3
Julia wants to publish her first article. How can she do it in a reproducible way?
Research Compendia
Organize files according to a prevailing convention.
Provide separation between data, methods and results expressing the relationship between the three.
Specify the environment
High reproducibility doesn’t mean that things couldn’t fail
What happens when the technology fails?
Some factors affecting Reproducibility
How the tools that ensure the reproducibility (Docker, Conda) change with time
Example 1 - Inactive Image Retention Policy (Nov 1, 2020): DockerHub removing Docker images in free accounts after 6 months
Some factors affecting Reproducibility
The software/hardware we use change with time
Example 2 - Mac M1 - Apple silicon chip (hardware) (Oct 2021) Docker issues
{fig.align:‘center’}
Software mantainance
Should our code have maintainers?
Ok, but I need a second career to learn all this!
1. Reprohack
Reproducibility hackathons
Authors: Send their papers with publicly available associated code and data
Participants: Attempt to reproduce published research of their choice from a list of proposed papers
At the end, the participants give feedback to the authors
2. The Turing Way project
The TWP involve and support a diverse community of contributors to make data science accessible, comprehensible and effective for everyone.
Book
3. rOpenSci / pyOpenSci
Scientific package review process
Software Carpentry
Lessons and infrastructure to teach them
The Carpentry lessons in Unix, python, R and more
research software engineer
What do RSEs do?
Fundamentally, RSEs build software to support scientific research. They generally don’t have research questions of their own. They develop the computer tools to help other people to do cool things.
Research Software Engineers includes researchers who spend a significant amount of time programming, full-time software engineers writing code to solve research problems, and those somewhere in-between.
US Research Software Sustainability Institute
US RSSI mission is to improve the recognition, development, and use , of software for a more sustainable research enterprise.
US Research software engineer association
Community-driven effort that brings together people who write and contribute to research software within the US.
US Research software engineer association
First US conference. US-RSE Conference 2023: Software Enabled Discovery and Beyond. October 16-18th, 2023, Chicago, IL
International RSE Conference. RSECon 23 September 5-7th, 2023, UK
What is the highest level of education you have attained? (one choice list)
How did you learn the skills you need to become an Research Software Engineer / Research Software Developer? (free text)
In which discipline is your highest academic qualification? (one choice list)
What training programs are you involved with? (free text)
Do you cite the software you use?
Software in general has not been well-cited (…), in part because the scholarly culture has not treated software as something that should be cited, or in some cases, even mentioned.
In general, when your software contributes to a paper, are you acknowledged in that paper? (one choice)
Publishing different article types
Data Papers
Methods Papers
Micropublishing
Software Papers
Registered Reports
Journal of Open Source Software
The Journal of Open Source Software is a developer friendly, open access journal for research software packages.
Master of Data Science program - UBC
10 month program - final capstone project with industry partners
Questions?
