and computing and science

are increasingly connected

"[The scientific discipline] concerned with constructing mathematical models and quantitative analysis techniques and using computers to analyze and solve scientific problems"

Methods

Web of Science

• Searched methods for "scientific computing"

• Recorded papers returned in even years across 15 disciplines

Increased computing

Increased women

Stagnant women in computer science

Case-paper search

- Searched computational science methods and "statistics"

- 2008-2012

- Retrieved 200 "most-relevant"

For each "case" paper, found "control"

- same journal, same time

Extracted all author first names

-Ran through gender classifier

Female authorship drops from 26.5% to 17.5%

ROpenSci analysis:

2. Extracted maintainer names for all active packages

1. Scraped CRAN

3. Ran through gender classifier

15% of ~4700 packages maintained by women

Alyssa Frazee (Johns Hopkins)

github

www.github.com

Code management system

~ 9 million users

Projects stored in repositories

Repositories belong to users (with user names)

Repositories map to primary language

Classified user names based on gender

Scraped username + language from all repositories with >= 5 stars

R

Male

Ambiguous

Female

Less than 2.5% of starred github repos map to female usernames

Students

Few mentors, limited guidance on CS skills

Science

BBSRC study of UK scientists: biggest future vulnerability is scientists' CS skills

Individual

ASA post-grad survey: computational skills are biggest liability

Institutions

Investment in infrastructure that people don't efficiently use

Few formal expectations

Culture ostracizes newcomers

Students enter with limited background

R is Open Source...

Bad track record wrt female/newcomer participation

~ 2.5% female; stable since ~ 2000 (... based on terrible data)

Data management

Code management

General hardware/software knowledge

Data visualization

Document preparation

(Data collection)

Doesn't know         doesn't care

Meet newcomers where they are

Encourage note-taking

 Explain why you do what you do

Point newcomers to other resources

1. Keep code clean and organized

Follow a style-guide (~10 mins of reading = much better code)

Use a consistent filing system

My ideas here:

http://ciddgsa.com/2015/04/11/a-few-steps-toward-better-cleaner-more-organized-code/

2. Self-Evaluate

Pick an old script to clean and functionalize

Assess where you're at wrt Data and Code management, basic logic ("Fizz-Buzz"),  Data visualization, Document prep

4. Expect to develop new skills

3. Simulate everything

Code Reviews

Peer reviews done once/semester (for credit??)

Collaborative projects like kaggle.com

• Data cleaning and prep
• Metric development
• Partition projects and design workflows
• Exposure to bigger data
• Builds scientific logic                              

Participate online

Twitter (#rstats, #ropensci), edit Wikipedia

Stack-overflow, r-help, etc.

Three major contributing factors

1. No clear expectations

2. Intimidating on-line community

3. Limited background skill

Articulate

Foster local community

Framework to catch up

There's a gender gap in scientific computing.

Lack of computing skill is problematic for science, individuals, institutions, and future workforce.