The Virtual Art of Statistical Consulting

Statistical consulting is an important but tricky business, and the life of a consulting statistician can be exhilarating, as this wonderful cartoon illustrates:

How not to collaborate with a biostatistician! Made with Xtranormal Movie Maker (not by me!)

Statistics’ graduates need to learn to become consultants. Existing programs typically include formal courses in statistical consulting, client problem seminars, placements with government, industry and partner institutions, and consulting in university statistical laboratories (Rustagi & Wolfe (1982), Cabrera & McDougall (2002)). The University of Melbourne offers all of these experiences to its students, through the Statistical Consulting Centre and the Centre for Data Science.

Undergraduate teaching of statistics all-too-often subscribes to the behavourist view of learning. In contrast, training graduates in the practice of statistics usually assemble the components above into an apprenticeship program, that is, they employ a social constructivist approach to learning. The benefits of this approach emphasize the human side of statistical consulting (Boen & Zahn (1982)). By encountering real problems brought by real people (authentic learning!), trainees learn to communicate with non-statisticians (Derr (2002)), write intelligible reports in language that clients can understand, admit their own ignorance, ask questions and listen!

Schuenemeyer (1990) stresses the importance of active learning (author’s original emphasis):

…The practice of statistics is something that you do… to train statistical consultants, we need to have them do consulting in a supervised environment…

Watts (1970) captures a similar sentiment (author’s original emphasis):

…Through such apprenticeships trainees may be exposed to the excitement and challenge of live problems. They realise for themselves the difficulty of formulating the statistical part of the problem, and they develop their own perspective of statistics…

The apprenticeship model, however, poses a problem in Australia. The vast majority of students who complete the Master of Biostatistics degree do so through the Biostatistics Collaboration of Australia (BCA), a nationwide consortium of six universities where subjects are taught entirely online. Despite this, one course objective states that graduates will “have developed the practical and technical skills to commence professional careers as independent biostatisticians”. Some students get experience pursuing the capstone subject “Workplace Project Portfolio” (required under the Australian Qualifications Framework for a Level 9 course) but most do not. Apart from channelling some students into NSW Health’s Biostatistics Training Program or the Rural Research Capacity Building Program (also in NSW), the BCA has been unable to address this competency problem.

One possible solution would leverage an important aspect of a face-to-face program: Trainees typically spend time observing and reporting on several consulting sessions (live or recorded), and practice being a consultant in a mock session. Participants are usually happy to have themselves video-recorded and then receive feedback from their peers and supervisor.

This suggests that bringing immersive reality (Cochrane et al. (2017, 2018, 2020)) to programs in statistical consulting could solve the problem of remoteness. It would allow trainees to, for example, visit a health services provider to observe their operations and data generation processes, sit in on a consultation between doctor and patient, play with medical devices, even attend surgery. I’ve done all of these things in person, but I had to work in a hospital for 10 years to do it! These are exactly the experiences that really matter to in the practice of statistics in an interdisciplinary framework (Lynch (2006)).

Aspiring consultant statisticians will be “job ready” sooner, competent and capable (Blaschke & Hase (2015, 2019)) to start working with clients who are “very busy”, think that they “know statistics” or who “just want an answer” – see the full list below in Figure 1. Perhaps we need a virtual scenario for each one!

Figure1: Extremes client types, their characteristics and the expected roles of the statistical consultant (by Gerald van Belle)

References

Blaschke LM, Hase S. (2019). Heutagogy and digital media networks: Setting students on the path to lifelong learning. Pacific Journal of Technology Enhanced Learning, 1(1), 1-14.

Blaschke LM, Hase S. (2015). Heutagogy, Technology, and Lifelong Learning for Professional and Part-Time Learners. In Dailey-Hebert A, Dennis KS. (Eds) (2015). Transformative Perspectives and Processes in Higher Education. Pages 75-94.

Boen JR, Zahn DA. (1982) The Human Side of Statistical Consulting. Lifetime Learning Publications.

Cabrera J, McDougall A. (2002) Statistical Consulting. Springer.

Cochrane T, Aiello S, Cook S, Aguayo C, Danobeitia C, Boncompte  G. (2018). Designing Immersive Mobile Mixed Reality for Paramedic Education. IEEE International Conference on Teaching, Assessment, and Learning for Engineering (TALE), University of Wollongong.

Cochrane T, Cook S, Aiello S, Christie D, Steagall M, Aguayo C. (2017). A DBR framework for designing mobile virtual reality learning environments. Australasian Journal of Educational Technology, 33(6), 54-68.

Cochrane T, Aiello S, Cook S, Aguayo C, Wilkinson N. (2020). MESH360: a framework for designing MMR-enhanced clinical simulations. Research in Learning Technology, 28: 2357.

Derr, J. (2000) Statistical Consulting: A Guide to Effective Communication. Duxbury Press.

Lynch (2006). It’s not easy being interdisciplinary. International Journal of Epidemiology. 35, 1119 – 1122.

Rustagi JS, Wolfe DA (Eds). (1982). Teaching of statistics and statistical consulting. Proceedings of a conference held at the Ohio State University. Academic Press, New York.

Schuenemeyer JH. (1990). Training statisticians to be consultants. Proceedings of the 3^rd International Conference on Teaching Statistics.  439 – 445.

Watts DG. (1970). A program for training statistical consultants. Technometrics. 12(4). 737 – 740.