Jun 12 2017

Feedback from the Cascadia R Conference participants

I helped organize the Cascadia R Conference on June 3, 2017.

About 190 people attended the all-day conference.  I volunteered to do the conference evaluation questionnaire and to analyze the results. We adapted a questionnaire that the CSV, Conf, 3 had used and used a Google Form to gather  feedback at the end of the day.  We got 59 responses or  32% of the conference participants, which seemed like an encouraging response rate.

The week after the conference, the organizing committee (which included Chester Ismay, Jessica Minnier, Lilly Winfree, Oliver Keyes, Scott Chamberlain, Ted Laderas, and me) chatted about the open-ended responses and discussed how things went in our Slack channel.   That actually seemed like an excellent, informal sense-making strategy for when an entire committee is data-oriented. This post combines some of their reflections (without specific attribution of specific comments or contributions from them) with a more systematic summarization of the data that I’ve done afterward.  Naturally I had to horse around with the response data in R and produce some graphs to depict what I thought was important. Ted Laderas Wrote up his reflections on the whole project in another post.

In scheduling a full day’s sessions – on a Saturday – one of our goals was to bridge across communities, specifically geographic communities (north and south along I-5).  We succeeded beyond our expectations.  About a third of the respondents came from more than 50 miles away with a bunch of people from more than 100 miles away.

Respondents asked for more social time: we know but probably always need to be reminded that R users are very social. (That’s an essential ingredient of R’s secret sauce.)  Definitely the day’s schedule was action packed, which we thought was a good thing. But, as one participant said, “The lightning talks were kind of rushed”,  as they were supposed to be.  When asked what their favorite thing was about the conference, 22 participants said “Workshops!”:

  • Workshops – 22
  • Lightning talks – 10
  • Meeting people – 7
  • Keynotes – 7

Two days might provide more social breathing room (for everyone except the organizers).  However a two-day format might be a challenge for people driving to Portland from far away.  In the future we could consider having a full “pre-conference” day for workshops and a separate day for talks. (We didn’t really know how popular the workshops would be.)  The useR conference uses the “pre-conference” structure for workshops.  They had 15 minutes for talks in 5 different locations with 3 minutes in between to transfer between talks so our setup for the talks was in line with their practice.

In the future we should just say upfront that this is not a traditional conference.  “To keep the cost of registration low we sacrifice <this> and <this> and <this>.”  And one of those is that you are on your own for housing.  A longer conference might tempt us to try to deal with an “official conference hotel”, but that would probably become a big headache.

Mimicking other conference formats and organizations like useR or the csv,conf is an important strategy for a small group like ours.  Afterward we noticed that the Open Source Bridge has the volunteer thing figured out, for example.  Some of us are going to that to pick up tips and strategies.

Here are the histograms from the new R skimr package, which suggests that everybody thought the location was great, a few people had problems with WIFI, and keynote topics and overall talk quality was very (but not completely) positive.

With really outstanding support from OHSU, rOpenSci, and Rstudio we did pretty well.

Interesting that we had a good mix of R mastery — from beginners to masters.  People who identified themselves as 3’s or 4’s were unusually enthusiastic about the keynote topics.

Respondents who said they thought participating in the conference would lead to future collaborations were more likely to say they would be willing to help organize (or volunteer) or that maybe they would be.  The cross-tabulation is below.  This pattern is clearer when you look at a Tukey median polish with residuals in italics and and the fit in bold.



Willing to help organize?


 Raw
data
Yes
Maybe
No

Met people
I'm likely 
to 
collaborate 
with
Yes
14
5
3

Maybe
5
14
12

No
1
2
2


Median
polish



Row
fit

Yes
10
0
-2
0

Maybe
-6
2
0
7

No
0
0
0
-3

Column fit
-1
0
0
5

Here is the code for fitting a median polish:

xtab <- xtabs(~ will_collab + help_org, data = feedback)
medpolish(xtab)

Also, respondents who said they were early or in the middle of their careers more frequently said they would be (or maybe would be) willing to volunteer or help organize a conference in the future.

Some selected comments from the open-ended responses:

  • Amazing conference! love the one day thing, the schedule, and the location!
  • I think the cheap cost and central location really contributed to the crowd size.
  • It would be great to have a keynote about ethics in data science.

Final suggestions seemed to be all over the place:

  • Less talks; more interactive activities
  • Longer
  • Longer breaks
  • More lightning talks, fewer full talks!
  • More talks and less workshops
  • Fewer, more in-depth talks
  • Explicit tracks + more than 2 tracks, maybe? “R in Bio” / “Stat Computing” / ???
  • Have three workshop levels next time – total beginner, intermediate, and advanced.
  • Notify the food carts ahead of time; they were really unprepared to be so slammed, and I think giving them a heads-up would’ve helped everyone

Finally, here are the topics we asked respondents:

  • Satisfaction with Registration
  • Satisfaction with Location/Meeting space
  • Satisfaction with Timing: Please rate the overall distribution of talk schedules and unstructured time
  • Satisfaction with Keynote topics
  • Satisfaction with Talk quality
  • Satisfaction with Conference agenda overall
  • Satisfaction with Quality of the wifi/internet access
  • Satisfaction with Snacks and drinks
  • Please add additional comments on the overall conference organization.
  • Where are you in your career?
  • On your way to R mastery, where are you now?
  • How many miles did you travel to get here?
  • Do you think you will form new collaborations as a result of attending?
  • Overall, the conference _________.
  • I considered the conference a _________.
  • I would like this conference to be held _________.
  • Are you able to help organize or volunteer next time?
  • If you’re able to help next time, please add your email.
  • What were your favorite parts of the conference?
  • What would you change or improve about the conference?
  • Please provide any additional comments below.

Being able to participate in the whole process, from hatching the idea to helping it actually happen, to participating in the whole conference and finally to thinking about what worked and what could be improved afterwards in this blog post was a great experience.  I’d happily do it again!

I’m even more convinced, if I needed to be, that the sociability and community-orientation that’s baked into R is profoundly important.  When we are struggling with a bit of code, gnarly data, or a graph that doesn’t quite look right, we tend to think of ourselves as working alone.  R provides many reminders that we are not alone and that data analysis is a social act.

No responses yet

Leave a Reply