Evaluation of data visualization effectiveness#

ArviZ is a Python and Julia library for exploratory analysis of Bayesian models, and includes many vizualizations to that end.

As ArviZ developers we strive to follow the literature on effective data visualization, especially around communicating uncertainty, but there are many factors that make this difficult. Including but not restricted to lack of time to implement or update code, publications that slipped under our radar, unconscious biases towards bad practices.

Therefore, to ensure ArviZ is an active agent in advocating and popularizing effective data visualization practices we budgeted $2000 as part of our GSoD 2021 application to pay someone external to the team to evaluate the plots that are available in ArviZ and their defaults.

Job description#

We are expecting a report covering the following points:

  • Analysis of available and missing visualizations

    Evaluate which visualizations should be present in ArviZ but are not. For each visualization we should get a description of how it would serve ArviZ’s goal and some degree of addition priority, not necessarily a ranking, could be binning missing visualizations in 3 or 4 high to low priority categories.

  • Analysis of current defaults

    Evaluate if ArviZ defaults follow best practices. Write down both things we are doing well (if any) and things we need to improve. Any and all points related to styling, layout, combination of multiple elements can be considered here.

  • Analysis of usage advise

    Evaluate current advise included in the ArviZ documentation about using each plotting function. As a first approximation this should be a list of useful references per plot so users can better understand when and how to use each of them.

The report should also consider the opened issues and PRs related to changes in our plotting module when relevant to each of the three points above.

We emphasize again the need for the report to include rationale for all comments and references where appropriate. While the report will be written around ArviZ, it will be published here and we expect it to be useful to other similar libraries like bayesplot.

Moreover, it could also act as a seed of a centralized resource maintained by all Bayesian data viz libraries with advise on how and when to use each plot, combining them…

Job requirements#

  • Not being part of the ArviZ team

  • Familiarity with data visualization and uncertainty visualization literature (TODO: any good way to evaluate this when getting applications?)

  • Familiarity with open source and ability to navigate ArviZ issues and documentation

Payment#

The budget for this project is $2000 which will be paid through NumFOCUS via OpenCollective, wire transfer or PayPal.

Don’t hesitate to contact us if you have any doubts about the payment process.

Contact#

For any inquiries about this project, please reach out on Gitter. If a question requires privacy you can also use the email below.

Applying for the job#

Applicants who wish to be hired to write the report should send an email to arviz.devs@gmail.com with the following information:

  • CV

  • Sample report

    We ask you to choose one of our plots and write a sample report for it regarding the two last points in the job description.

  • Statement of interest

    Short document (between half and a single page) explaining why you are interested in writing the report and why do you think we should hire you. Make sure to emphasize how you meet the job requirements (described above): publications in data viz, involvement with open source projects…

  • Tentative Work Plan and timeline

    Outline of your planned work. This should include work content and a timeline.