International Seminar on Statistics with R

One of the first to spot the demand was Luciane Alcoforado, a professor at the Fluminense Federal University (UFF) in Rio de Janeiro, Brazil. Realizing there was no large R event in Brazil, she set about coordinating the first International Seminar on Statistics with R (SER: Seminário Internacional de Estatística com R), which was held in Niterói in May 2016.

Luciane Alcoforado, Coordinator of the first SER

More than 250 people attended this pioneering event, with attendance balanced across academia and industry. Encouraged by this success, the I SER organizers ran the event again in 2017, with an expanded program and similar numbers attending.

Sean Kross presents remotely at the first SER

A third edition, III SER, is now planned for May 22-24, 2018, to be hosted again by UFF in Niterói, Brazil. On the first day, 8 mini courses will be run, aimed at both practitioners and students. This year most of the international guests come from the USA and England. They include Jesse Mostipak, Becky Pattinson, Garreth Maccray, and Gabriela de Queiroz. These guests will present alongside invited speakers from the local region. There will also be thirty selected contributed talks and a poster session with awards for the best three of each. It’s a huge opportunity to get your research known. The last conference day will close with a tribute session to Prof. Djalma Pessoa who was a pioneer R user in Brazil.

Registration to III SER and the mini-courses is open now and the call for papers is open until March 23 2018.

R Day

A second event in Brazil, R Day, will take place at the Federal University of Paraná (UFPR), Curitiba, on May 22, 2018. The Statistics Department at UFPR was one of the first in Brazil to adopt R as the standard programming language for statistics teaching, in 2002. It was also the first institution to host a CRAN mirror in the country, and started the R-br mailing list in 2011, which now has more than 1000 subscribers.

The R Day is organized as a satellite event of the 63rd RBRAS meeting (Brazilian Region of the International Biometric Society), and is endorsed by the R Foundation to further develop the academic R community in Brazil. There will be three keynotes, 9 oral presentations, and 8 tutorials, aimed at both users and developers. The main goal is to be community-driven, so anyone can submit proposals for oral presentations and/or tutorials (only the three keynote speakers will be invited by the organizers).

The aim of R Day is to gather the vast R community in Brazil, by offering the possibility for users and developers to show their work and exchange experience on all sorts of R subjects. Proposals for oral presentations and tutorials are open until April 04. The registration is free and can be made until April 22.

For more information and updates, registration and proposal forms, checkout the R Day website, and follow us on Facebook and Twitter.

R Day logo (left), Curitiba Botanical Garden (top right), and UFPR historical building (bottom right)

satRday Bogota

The first Latin American chapter of the one day conference series satRdays will take place in Bogota, Colombia in August 2018. The focus of the conference will be about tidy concepts applied to data visualization. It is organized by Datasketch, a Data Journalism Portal for Latin America.

The main goal of the conference is to create a common space for interaction among newbies and experts in R. Special interest will be given to works created by or for underrepresented communities and to showcase different solutions developed using R for the different elections that are happening in 2018 in different Latin American countries.

There are tentative plans for a workshop session the day before with hands-on tutorials about:

• Data visualization and common mistakes.
• Data structuring and preparation: PDFs, Databases and API’s.

A website containing more details will be available soon.

LatinR

On October 25, 2017, an announcement was made by Heather Turner on the RUG Organizers Slack: the R Foundation Conference Committee would like to see academic-focused R events in regions not currently covered by useR! In less than a week, a group of Latam R-Ladies were organizing their first conference-call to start thinking about how this challenge could be achieved. This quick response was not the result of mere chance, but the consequence of a year in which the R community — both RUG and R-Ladies groups — grew stronger in South America. By mid November everything was set-up: a name, a place, a date, and a motivated international organizing committee.

LatinR 2018 will be held in Buenos Aires on September 4-5 within the 47th Argentinian Meeting of Informatics and Operational Research (JAIIO), which is organized by The Argentinian Informatics Society (SADIO). With a diverse and renowned Scientific Committee, this two-day event aims to bring together researchers, students, and R-enthusiasts from different backgrounds in order to strengthen the R community and make the impact that R is having in the region more visible, in both academic and non academic contexts.

For more information about oral presentations & poster submission, keynotes, and workshops, you can check out the LatinR website or follow us on Twitter and Facebook: @LatinR2018.

ConectaR 2019

In 2019, it is Costa Rica’s turn: ConectaR 2019 will take place in San José. The event has the endorsement of R Foundation and will be hosted at the University of Costa Rica (UCR). ConectaR has an organizing committee comprising members of academia (School of Statistics, Pure and Applied Mathematics Research Center (CIMPA)), public research centers (The Development Observatory, Bioversity International (CGIAR)) and industry (ixpantia). ConectaR will have two keynote speakers (yet to be confirmed) and will be receiving proposals for oral presentations and posters starting April 2018. In addition to that, there will be tutorials, invited talks, and a job fair. An attendance of over 150 people is expected, with the goal of building a community of R users in the Central American Region.

R/Bioconductor Latin American Developers’ Workshop

In recent years, biology has seen a rise in the use of technologies that enable high-throughput, quantitative, data-rich profiling of cellular states. As a result, the field faces computational challenges to analyse such data. The R/Bioconductor project is an open source, open development software platform that provides tools to translate complex data sets into biological knowledge. An R/Bioconductor workshop will be held in Cuernavaca, Mexico on July 30 - Aug 3, 2018.

The workshop is aimed at students and researchers interested in the analysis of biological data. We encourage applications from diverse disciplines, including but not limited to biologists, bioinformaticians, data scientists, software engineers and programmers. The main goals of the workshop are:

1. Teaching participants the principles of reproducible data science through the development of R/Bioconductor packages.
2. Turning bioinformatic software users into bioinformatic software developers.
3. Fostering the exchange of expertise and establishing multidisciplinary collaborations.
4. Creating a community of Latin American scientists committed to the development of software and computational pipelines for biological data analysis.
5. Helping to train bioinformatics instructors that can continue to grow in their local communities.

This workshop is part of a long-term project to create a community of developers from Latin America. We hope to hold regular meetings in the future (similar to BioC, EuroBioc and BioCAsia) where attendees present their own software contributions.

A website containing more details will be available soon.

Local groups

Beyond conferences and workshops, local groups provide a venue for R users to meet more regularly. Jumping Rivers provide a user group directory, which also covers R-Ladies groups, however as R-Ladies are expanding rapidly it is worth checking the R-Ladies Starter Kit for the latest status.

Recently members of the RUG and R-Ladies groups in Bogota, Columbia collaborated on a striking intervention about feminicides in Colombia using plots created in R. Read more on their initiative in this report (warning: readers may find the content distressing). This is an example of how local groups can work together on wider issues as well as providing networking and development opportunities for their members.

Making Connections

Part of building the community in Latin America is making connections between disconnected groups. If you think we have missed something, or you would like to be involved in plans for future events in the region, let us know in the comments or contact us on email or Twitter.

Banner image shows the poster session at I SER.

Belonging to the Community

The first question asked useRs whether they considered themselves to be part of the R community: 361 (79%) said yes, 69 (15%) said no and 28 (6%) did not give an answer. There was no significant difference in responses from men and women.

It’s encouraging that such a large proportion feel part of the R community: as might be expected the proportion increased with the length of time the respondent had used R, yet even people with less than 6 months’ experience responded positively. This is a testament to a generally welcoming community as backed up by some of the free text comments, for example:

Y’all seem to have a very active and robust community - keep it up!

Support in Using R

The next question in the community section asked useRs which resources they used for support. Respondents could select one or more resources from a list and/or add their own resources. The responses were comparable for both men and women, so the plot below summarises the results for all respondents, showing the resources that were selected by at least 10 individuals:

Over 70% (334 of 458) use StackOverflow for support, but the long-running R mailing lists are still heavily used, with 40% (180 of 458) using these. Both resources were criticised in the comments as being unfriendly to newcomers though, for example:

The mailing lists need to be more inclusive / friendly. Way too often new users are scared away by the “read documentation” and/or “this is a statistics question” comments

Getting R News

Next we asked useRs what was their preferred medium for R community news (e.g. events, webinars, opportunities). Again there was no significant difference between the responses for men and women, so the results for all genders are presented below:

The majority of useRs are split across four media: website (24%, 110 of 458), mailing list (21%, 96 of 458), blog (21%, 85 of 458) and Twitter (15%, 69 of 458). Therefore communication across a range of media is necessary to be confident in reaching a large proportion of R users and developers. In addition to the other media featured in the plot above (Facebook, RSS feed, podcast) useRs selected Google+ and their local R meetup group as sources of news. The remaining people in the “Other” category wanted to select multiple categories or had no preference. An exploratory multivariate analysis¹ suggests that Facebook (not Twitter), podcasts and blogs are more more popular among newcomers, especially women, so these media may be important for outreach. A sizeable proportion (13.1%) did not select any option, so do not have a preference or are not interested in R community news. Nonetheless, as conference participants, they may receive some news in person.

Local Groups

The last set of directed questions related to useR’s involvement in local R user groups. Among the 458 respondents, 152 (33%) said they attended a local group, whereas 279 (59%) did not and 27 (6%) did not respond. Of those that attended RUGs, 124 (82%) attended a general user group, while 12 (8%) attended a university-based group and 7 (5%) attended a company-based group. Other types of groups mentioned were a user group for women in R (R-Ladies), a combined general user and R-Ladies group, a group based in a government organization and an informal user group for government and academia.

The respondents that did not attend a user group were asked why not, giving the following responses:

The most common responses were that there was no active group nearby (30%, 85 of 279) or they were too busy (30%, 84 of 279). The next most frequent responses correspond to people that feel too inexperienced (11%, 30 of 279) or perhaps too experienced, in that they don’t need the extra support (9%, 25 of 279). Some people did not know whether there was a group near them, or were unaware that such groups existed - one person looked up their local group as a result of filling in the survey and signed up for it straight away! For others the topics were not relevant/interesting, the time or place was inconvenient, they didn’t use R enough (using other tools more or being completely new to R), there were logistical issues (accessibility, problems signing up to meetups or not enough space at meetup) or social issues (feeling intimidated or worrying they would intimidate others), they’d just never considered it, or planned to go soon.

A higher proportion of men (46%) said that they attended a user group, compared to women (25%). Men and women also differed in some of their responses when asked what would make them more likely to participate in the R community, or improve their experience. Therefore these results are shown for men and women separately below, for options with at least 5 respondents in a gender:category combination:

The top two options were to have a new user group nearby (27%, 122 of 458) and mentoring (19%, 87 of 458). The proportions selecting these options were similar for men and women. For women, the next most wanted support was an online forum to discuss R-related issues; 20% of women vs 10% of men selected this option. The gender difference was similar regarding views on an online support group for their demographic (note this could be other than gender), but overall the positive response was much lower for this option. For men, after user groups and mentoring, the next most popular option was workshops, with similar numbers selecting free local introductory workshops, paid local advanced workshops and workshops at a conference in their domain. These options were also popular with women, with a slightly stronger preference for advanced workshops or workshops at conferences. Conferences specified in the comments were JSM, INFORMS, NIPS and ICML. For men, there were no other options selected by at least 10 individuals, but for women there was still support for user groups, workshops and an online support group aimed at their demographic (respectively 13%, 8% and 8% of women). Respondents were asked to specify the demographic in the comments, mostly this was women, but LGBT and academics (students/teachers/reseachers) were also specified.

Less frequently selected options, or aspects mentioned in the comments were childcare at conferences, a user group where it is possible to bring children, greater demand for R in local area and training in a non-English language (Chinese).

The community maintains a list of R User Groups and R-Ladies Groups can be found there as well as on the R Ladies Global website. As the conference was in Stanford, it is not surprising that there were several people from the local area interested a group near them. For some of these, the Bay Area R User Group, might work, but some individuals e.g. from Oakland or Palo Alto felt this was too far. Further afield there were a number of specific places mentioned, where there is already a local group or one not so far away, though many have started since useR! 2016:

• Los Angeles, CA RUG
• Denver, CO RUG
• Connecticut CT R-ladies (NEW)
• Washington DC RUG; R-Ladies (NEW)
• Miami, Florida R-Ladies (NEW)
• Orlando, Florida R-Ladies(NEW)
• Ames, IA R-Ladies (NEW)
• Boston, MA R-Ladies (NEW)
• Greensboro, NC; not so far from Winston Salem RUG
• Durham, NC R-Ladies (NEW)
• Columbus, OH R-Ladies (NEW)
• Portland OR RUG
• Nashville, TN RUG; R-Ladies (NEW)
• Seattle, WA RUG; R-Ladies (COMING SOON)
• Zurich, Switzerland RUG
• Muenster, Germany RUG (NEW)

There were also some places where there isn’t an active group nearby and there may be potential to (re)start a new group:

• Central Valley, CA
• Davenport, IA
• Iowa City, IA
• SE Minnesota, MN
• Wilmington, NC
• Moscow, ID
• Eugene, OR
• Richland, WA
• Geneva, Switzerland
• Hasselt, Belgium
• Freiberg, Germany
• Glasgow, Scotland
• Cluj-Napoca, Romania

Many domains were mentioned including spatial statistics, electrical engineering, operations research, health care and education. Some of these might be fed through to the corresponding local groups, while others might provide ideas for future useR! conferences.

General Comments

The survey concluded with two open-ended questions, first asking respondents for their ideas for improving the R community and second for general feedback on the survey. Some of the comments we have mentioned earlier, specific ideas from the remaining comments are outlined below.

Community/R Project organisation:

• A webpage introducing the community and how it is organised online
• Taskforce for communities/local groups
• Promote local groups
• Move R core development and CRAN to a more inclusive platform such as GitHub.
• More women in leadership (R Core, R Consortium, R Foundation)

useR!/conferences:

• A special gathering for first-time attendees at useR!
• A beginners’ track at useR!
• Workshops at useR! to help people solves problems they bring along
• Childcare at useR!
• Informal “birds of a feather” tables at conferences
• Continue with the diversity scholarship to help increase diversity at useR!

Local groups:

• Publish guide for running local group
• Meetings aimed at different levels of expertise/experience
• Team participation in competitions, e.g. kaggle
• Voluntary work on data science problems for non-profits
• Open source sprint, e.g. on R tools for sight impaired
• Open source introductory materials for groups to run workshops
• User group [events] aimed at users who want to contribute to packages/base R
• Joint events with other data science groups
• Collaborative projects across user groups
• Online virtual R conference among user groups
• Global competitions to resolve core R issues

Professional development/education:

• Slides repository on git like the Bioconductor courses resources
• R curriculum for undergraduate courses
• Guideline for switching from other languages, e.g. SPSS
• Resources to help long-time useRs keep up-to-date (try hipsteR!)
• Mentorship, particularly for first CRAN submission
• Encourage networking between established and new useRs
• More workshops

There was also a call for more outreach, particularly to young people (especially girls) in high schools, but also to educators, social scientists and business schools.

There is a lot to work on here, but several respondents commented on their motivation to contribute, particularly after attending useR! and hearing of the work of Forwards (or the Women in R Taskforce as it was then) and R-Ladies. Others said they didn’t have the time to contribute very much, but in some cases would be prepared to contribute if they were asked to do something specific.

Finally many respondents expressed thanks to Forwards for running the survey and for our efforts to improve diversity and inclusion - many thanks for your encouragement!

Conclusions

It has taken us a while to work through the survey and report back. However a benefit of this is we can see that actions in the meantime by Forwards and the wider community are already starting to address many of the issues raised in the last part of the survey.

The massive expansion of R-Ladies over the past year (from 4 groups to over 45²) means that many people who were looking for a group now have one nearby, or will have soon. Many of the suggestions for useR! are in place for useR! 2017: there will be a newbies session, childcare and a “hackathon space” for informal birds of a feather gatherings. In addition 26 participants were awarded diversity scholarships (thanks to the conference sponsors, especially platinum sponsors Microsoft, Open Analytics, R Consortium, RStudio, DataCamp, and IBM).

With regards to education and outreach, the Forwards Teaching sub-team were recently awarded an R Consortium grant to develop workshops specifically targeting women and high school girls - initial workshops are planned for September and December in Melbourne, Australia and Auckland, New Zealand respectively. Further workshops will be run in Europe and the US, and the material made available under a Creative Commons share-alike license for others to use.

Beyond this, we can use the ideas and findings here to inform future initiatives. For example the Forwards Community sub-team can help to promote local groups by raising awareness, connecting people that want to start new groups and sharing ideas to increase participation. In line with our mission to increase participation among under-represented groups, we can focus on initiatives that appear to be particularly appealing to women, such as mentoring and an online forum.

Overall we have learnt a lot from our survey of useR! participants. In future we would like to study other populations and have package authors in mind for our next big survey. However, we still wish to monitor participation of under-represented groups at useR! so will be running a demographic survey as part of the general conference evaluation survey at useR! 2017. If you’re coming to useR! please support us by responding!

Notes

Now that we have completed this blog series, we will work on preparing an aggregated data set that minimises the risk of disclosure. We hope to release this over the summer.

The analysis presented here used the following packages:

• For data manipulation: dplyr, forcats, tidyr
• For visualization and reporting: knitr, ggplot2

The code is available in the Rmarkdown file.

The banner photo was taken at useR! 2016 by Thomas Petzoldt, used with permission.

Programming experience

We first asked attendees how long they had been using R for. Unsurprisingly for attendees of a conference all about R, the respondents tend to be rather experienced, with more than half (57%, 251 of 442) using R for at least 5 years. Nonetheless, the conference does attract people new to R, with a small number having less than 6 months experience. A greater proportion of women than men have fewer than two years experience (24%, 39 of 164 vs 12%, 34 of 276) which correlates with the younger age distribution we noted in our first post. The following plot shows the percentages within each experience category for each gender:

66% of women (109 of 165) said they had programming experience before learning to use R, compared with 82% of men (227 of 276). Since women tend to be younger, this could be due to increasing use of R in teaching, particularly on undergraduate courses. Thus younger people generally are less likely to have used other languages for data analysis such as SAS, Stata, S-plus, Genstat, etc. However it could indicate a different background of women vs. men, in particular they may be less likely to have any formal training in computer science, where they might have been taught C/C++, Java etc before learning R.

We also asked about how people use the language. Respondents were asked to confirm that any or all of the following statements applied to them:

• “I use functions from existing R packages to analyze data”
• “I write R code designed to make my work easier, such as loops or conditionals or functions”
• “I write R functions for use by myself or my collaborators”
• “I contribute to R packages (on CRAN or elsewhere)”
• “I have written my own R package”
• “I have written my own R package and released it on CRAN or Bioconductor (or shared it on GitHub, R-Forge or similar platforms)”

As might be expected given the length of time useR! attendees have been using R, their use is generally quite advanced with the vast majority using techniques to control the flow of their code (86%, 392 of 458) or writing custom R functions (85%, 389 of 458). More than half have written their own package(s) (52%, 238 of 458) and a third have publicly released a package (34%, 155 of 458).

Reflecting the distribution of length of R programming experience, the proportion of women using advanced programming techniques is slightly lower than men and when it comes to package development the proportion is lower still:

The final question we asked regarding use of R was whether the attendees use R primarily for work, recreation or both. The results were significantly different for men and women, only 22% of women (35 of 162) use R for both work and recreation compared to 45% of men (147 of 275); the few people that use R for recreation only were all men. Although factors such as age and programming experience may come into play here, it is likely that social expectations of women are also a factor. For example, although responsibility for cooking, childcare and housework is shared more evenly among working couples, in different sex couples women often shoulder a greater responsibility¹. With reduced leisure time, women may be less willing to use R just for fun, or may feel less able to do so, due to expectations of their family and friends about how they should spend their free time.

Overall, this paints a picture of useR! attendees as a population of mostly long-term R programmers often writing code for more than their individual use - which makes sense. As we discussed in our first post, conferences are expensive in terms of both time and money, so a conference on R is most viable for people that use R substantially in their work and hence can justify the cost to their employers. As the main conference organised by the R Foundation, useR! is particularly attractive to dedicated contributors to the R project, indeed talks and posters should present a contribution which will often be a new package or contribution to a package. This goes some way to explaining some of the inequalities in attendee demographics; we know from separate research (albeit using a rather crude metric) that CRAN maintainers, for example, are mostly not women.²

We have seen that the women that do attend useR! tend to be younger and less experienced. We have not explored the programming experience of other under-represented groups, for example with regard to ethnicity or sexual orientation, as the numbers in the sample are much lower, prohibiting reliable statistical conclusions and creating issues with identifiability. It is possible that, like women, members of other under-represented groups are also less likely to use R for recreation, as they will tend to have less time or money than is the norm.³ This indicates a barrier to women and other under-represented groups transitioning from useRs to developeRs: package authorship and maintenance, and dedication to an open source community, is largely uncompensated.

Opinions about R

The second part of our section on R programming asked respondents for their opinions on R. On these questions there were no significant differences between the responses from men and women, so we present the results for all respondents, including non-binary individuals and those that did not declare their gender.

We first asked respondents how strongly they agreed or disagreed with the following statements:

• writing R is fun
• writing R is considered cool or interesting by my peers
• writing R is a monotonous task
• writing R is difficult

The results are presented below:

We see that respondents overwhelmingly agree that writing R is fun, so even those people that use R primarily for work enjoy doing so. The view of their peers is slightly less positive, with 12% (51 of 438) stating that writing R is not considered cool or interesting by their peers and 15% having no opinion on this (65 of 438). Even among those who agreed with the second statement, the proportion that strongly agreed is lower compared to the first statement. So although the overall picture is positive, there is a sizable group for which writing R is not encouraged by their peers, which could affect their motivation to contribute, particularly outside of work.

The majority (61%, 269 of 438) disagree with the statement that writing R is difficult. Those that agreed with this statement are spread across the experience levels, so some may have been expressing a general opinion about R programming rather than a personal one, necessarily. The disagreement is stronger (80%, 349 of 438) against the statement that writing R is a monotonous task, so while writing R may occasionally be difficult most useR! attendees find it challenging and interesting.

The generally positive view of R among useR! attendees is not so surprising, after all most will have chosen to attend the conference, or if it was required by their work they have expressed an interest in R by applying for that job or choosing to use R at work in the first place. Nonetheless, the positive view is a benefit to the community as useRs encourage people to take up R and contribute more to the R project, with many useRs being in a position to teach R, mentor junior colleagues/students or contribute to community events such as R user groups or R-ladies.

91% (418 of 439) said that they would recommend R to friends and colleagues. The reasons some people would not recommend R include a steep learning curve, too many idiosyncrasies, poor performance and that it encourages bad programming practice. Some simply said they would recommend other software over R or they may recommend R, depending on why their friend or colleague wanted to learn a programming language.

When asked to pick a single reason for recommending R, nearly 40% (160 of 425) said it was good for statistical analysis. R has long been described as the lingua franca of statistics or data science, so this makes sense. The next biggest reason was that it was good for reproducible research, indicating the increasing importance that has been put on this topic in recent years. Other important features are the package ecosystem, strength at visualization and the fact that it is free and open source.

Outreach to under-represented groups

As the R community seeks to be more inclusive of under-represented groups we need to be aware that individuals from these groups are less likely to have a programming background and may have less freedom to spend time on unpaid and recreational contributions. Given R’s strength as a tool for statistical analysis, introductions to R should focus on this rather than the nuts and bolts of the language, as a general rule. More experienced R programmers from under-represented groups may need additional support to transition to developeRs – even when writing packages in pure R, concepts such as version control and object-oriented programming may be new and of course interfacing with other languages such as C++ is a further stretch. We should seek to promote opportunities to under-represented groups that are paid and/or will benefit their career, e.g. Google Summer of Code projects, useR! diversity scholarships, professional development via R-ladies groups, developing a package with rOpenSci, publishing in The R Journal and so on.

Taskforce initiatives

Following our recent call to join Forwards we have set up a number of sub-teams to support women and other under-represented groups in the R community. In particular the teaching team will focus on more inclusive pedagogy, such as using examples to engage a wider demographic, and running workshops for women and girls. We have also set up a community team, to look at ways to reach out to under-represented groups and an on-ramps team to look at ways to help useRs transition to developeRs. These are all new initiatives, but we hope to get some projects off the ground in all these areas in the coming year. Anyone wishing to get involved can contact us at rowforwards@gmail.com.

Further research

Some interesting questions for further research on useRs relationship with R include:

• What programming experience do people have before using R? For example, what languages did people learn before R?
• Why do people write packages but not publicly release them? For example, are the packages for their own use (work/recreation), for internal company use, or are the packages in development? One respondent commented that their employer prohibited releasing packages on CRAN; how widespread is this?
• What are the pros and cons of using R? Several respondents commented that R has many benefits but also some drawbacks and richer data could be collected on this topic.

Gender and race

The gender breakdown of responses (n=453) showed that approximately 33% of attendees identified as women, with the majority of the remainder identifying as men and a small number identifying outside the gender binary.

While the inclusion of women is not as bad as at other conferences, it’s also definitely lower than it could be given that there are many scientific fields - such as biology, which R is particularly commonly used in - where women outnumber men in terms of degree attainment.² At the same time, some of this may be offset by the location and chosen topics for presentations, which were more focused on visualisation or computing than biostatistics.

Race is more complicated, because there’s more variation in the answers - as intended! Both race and gender questions were offered as a free text field to avoid restricting respondents to false dichotomies. One side-effect, though, is that it’s hard to aggregate the data. What we can say, though, is:

• 74% of respondents identified as White or Caucasian;
• 16% fell within the US Federal guidelines for ‘Asian’, or identified directly as Asian;
• 3% identified as Hispanic or Latinx;
• 1% identified as people of African descent;
• 0.5% identified as Native American or Pacific Islander;
• 0.25% identified as Middle Eastern and
• 3% identified as mixed-race.

In comparison, a survey of STEM workers in the United States found 6% were Black and 7% Hispanic^{3 4} - and the overall demographics of the United States point even more clearly to an under-representation of non-Asian minority groups.⁵ The response rate was somewhat lower on this question (n = 409), perhaps due to concerns about identifiability. Nonetheless, there’s clearly work to be done.

LGBT

5% of respondents identified as Lesbian, Gay, Bisexual or Transgender. How this maps to overall industry, academic and population trends is hard to identify, because the long history of persecution of LGBT individuals and the often confusing and ill-defined nature of identity means that the proportion of LGBT individuals is regularly under-reported. Attempts to control for that under-reporting have shown up to 19% of the U.S. population experience same-sex attraction,⁶ although more traditional methodologies suggest around 11%.⁷

It’s possible that the useR! survey is one of the surveys in which under-reporting is found - although 447 responses were received, which suggests people at least didn’t skip it - but it’s probably fair to say that such a sizeable difference in outcome is going to contain some genuine differences in population, as well as methodological variation.

Additionally, it is probable that the other demographics of LGBT attendees reflect the overall demographics of attendees, meaning that (for example) we would expect to find lesbians or bisexual women under-represented in that 5%.

Age

Approximating age (by subtracting the year of birth from 2016) shows the following:

Women trend younger than men at the conference, but the age distribution is similar across the two genders; 47% of respondents (n = 438) were born in the 1980s - in other words, were people in their late 20s to mid 30s. This lines up neatly with the education data (more on that later) and suggests that people in their early 20s and 40+ year olds are not finding the conference appealing - because there are other things to spend time on, because money is tighter, or for some other reason.

The women-being-younger element suggests that the rise in women attending (compared to 2015, where the proportion of women was only 19%) represents younger or newer members of the community attending, too.

Geography

The R community is a highly international one, with developers in a wide range of countries. One of the things we were interested to discover is how this comes out in the attendee mix at useR!. Accordingly, we asked attendees which country they currently resided in.

Altogether, R programmers from 34 nations attended useR!, but the overwhelming majority of respondents (71% of n = 443) live in the United States. In some ways this isn’t surprising - the cost of getting to Stanford increases the further away you are, and it’s expensive to start with! - and the rotation of the conference between continents each year means that many non-US R programmers are likely to simply attend the next European event. At the same time, it’s still somewhat disappointing for an event considered the “main” R conference, particularly with only ~20 attendees coming from Canada, a country bordering the U.S.

Education

The vast majority of attendees (>85%, of n = 451) had completed postgraduate degrees:

This further suggests, on top of the data around age, that the conference is not attracting younger and more-early-career R programmers and researchers. Part of this may simply be that spending money on an R conference may be a lot easier to justify when you’re in full-time employment, or in a postgraduate degree with a stipend, than as an undergraduate or masters student; part of it is probably also that people without degrees, or without postgraduate degrees, earn less money and may come from poorer backgrounds, extending the problem.

At the same time, it is worth noting that the concept of research conferences as places to attend is not something particularly pushed on undergraduates, and so there is a component of awareness as well as finance. Finally, it may simply be that the conference is very academia-oriented, and so is particularly attractive to academics (and those familiar with academia) rather than R programmers without an academic background.

As the age data suggests, women tend to be slightly earlier on in the traditional academic career path.

Professional environment

Almost all of the attendees were employed, which (with the cost of the tickets) makes sense. More interestingly, a majority of the attendees (~57% of n = 444) were employed outside academia, primarily by private sector organisations. This is a strong indicator of what we’ve known anecdotally for a while - that R, a language originally designed and used in academia, is becoming more prevalent outside that context. At the same time, this may be a (somewhat) localised effect: while the language is increasingly popular, the technology industry centred on Silicon Valley is certainly an outlier in many respects for research overall.

Another interesting takeaway was that men are more likely to work in industry and permanent academic jobs than women - and women are far, far more likely to be students. Again, this isn’t surprising since it fits the pattern that women attending tend to be younger than men, and overall patterns that the tech industry (which is where one would expect the majority of privately-employed R programmers and scientists to work) has a known problem with employing and promoting women, as does academia.

Almost all attendees work full-time. It’s difficult to tell if this is due to the wider community largely being employed or engaged in full-time activities, or due to the cost of attendance diminishing the ability of community members in more economically fragile situations such as part-time work or unemployment to attend. Female attendees were less likely to work part-time, and far more likely to identify their status as unemployed (although the sample size is too small to really draw conclusions):

The proportion identifying as unemployed is greater than the proportion that previously identified as unemployed or retired, which could indicate that some students are categorising themselves as ‘not in employment’ when not presented with an option that reflects student-status.

Caregiving

Approximately a quarter of respondents answered “yes” to the question “are you a caregiver for children or adult dependents on a regular basis?” (n = 445). More men than women identified as caregivers. This either indicates that R programmers are a singularly socially enlightened population, or reflects wider sociological conditioning around caregiving which makes it easier for men to leave children at home with their partner (if applicable) than for women to do the same - resulting in a self-selection effect in which women with children are less likely to be able to attend (and so women who do attend are less likely to have children they care for).

Qualitative data suggests, unsurprisingly, the latter - in the free-text field for the survey, several (female) respondents mentioned the difficulty of persuading their (male) partners to handle childcare to allow them to attend conferences.

Conference proposals

The survey data surfaces a lot of problems that can be tackled directly by future useR! conferences, and other conferences in the field too. In particular, it shows that while the gender breakdown of the conference is not as bad as we might fear, it’s still lagging behind the general population of the scientific fields R programming draws from, as well as the population as a whole. In addition, and in some ways more worryingly, attendees are overwhelmingly white, with Native Americans, Latinx people and people of African descent particularly under-represented.

These problems won’t be solved overnight, but there are several things the conference can do that would make attending a more viable and enjoyable prospect for under-represented groups. These include:

1. A wider scholarship system. useR! 2016 featured scholarships, helping pay for things like plane tickets or hotels for people from marginalised backgrounds who wished to attend. The committee tasked with allotting the money did excellent work, but the money available was small. Expanding the scholarship fund in future conferences would let more under-represented R programmers attend, going a long way towards making a more representative conference.

2. Childcare and feeding support. One of the things we saw in qualitative data, and something the quantitative data points to, is that the difficulties of juggling childcare is a blocker on attendance, particularly for women. Providing subsidised or free childcare at the conference itself (along with breastfeeding rooms), and incorporating childcare into the scholarship system, would help with this, and has been successfully done elsewhere.⁸

3. Staggered ticket release. useR! 2016 sold out incredibly quickly, with all of the tickets gone within a day. This made it difficult for people to attend if they weren’t extremely on the ball and able to decide, on or before the first registration day, that they wanted to attend. By definition this hurts marginalised people because it hurts people who, for financial reasons, need to put some energy behind working out if they can attend. One solution worth considering is staggering the release of tickets over several days or weeks - perhaps with a waiting or notification list. This offers potential attendees multiple opportunities, reducing the penalty paid for not being ready to go on day 1.

4. A more precise code of conduct. Following an open letter in 2015 the R Foundation adopted a resolution that any Foundation-affiliated conference have a Code of Conduct,⁹ which we applaud. What that code contains, however, in terms of restrictions, the amount of information given to attendees, and the enforcement mechanisms and their ease of use, is largely left up to the individual conference instance. This creates inconsistency and uncertainty for both attendees and conference organisers. Selecting a specific code of conduct, with a specific enforcement mechanism, along with training for enforcers, would solve for much of this, and give concerned attendees more certainty that the conference is a safe place to be.

Task Force proposals

This survey has also helped to shape the goals of the Task Force on Women in the R Community. In particular, focusing in women, as a catchall, is not enough. Issues such as race, sexuality, disability and class intersect with gender and must also be considered if we are to build an R community that is more representative of the scientific/software communities and of society as a whole.

Accordingly we are rebranding the Task Force as simply Forwards, so that we can consider the needs of other marginalised groups as part of our mission. Our primary focus will still be on women as there remains much work to do here and broadening participation of women was the motivation for setting up the task force. However many of the practical actions we can take, such as the proposals above, will also impact other marginalised groups and making this impact a formal part of our mission will help to ensure our actions do not only benefit white, highly educated women.

As an immediate consequence, we will be publishing an open call for participation, seeking more members for the group. We are particularly interested in talking to volunteers from non-White, LGBT, non-middle-class or early career backgrounds.

Running this survey has also served to reinforce that we have no fundamental ground truth here; it is hard to see what the backgrounds of those who try but fail to attend useR are, and their reasons for not going - it is hard to see what the demographics of R community members overall look like. Accordingly we will be running further surveys, adjusting several questions to factor in things we’ve learned from running this one, looking at the overall community and useR! attendees in a more nuanced way.