Many have written and spoken about data science as a team sport. We hear stories from scientists, engineers, researchers, and statisticians about working in teams. Sometimes, it’s inspiring to hear about these beautifully accomplished collaborations. Other times, it’s painful to hear about frustrations, misunderstandings, and resource-wasting mistakes leading to missed insights or unanswered questions.
Having taken the excellent American Statistical Association Ann Arbor Chapter short course, Navigating Tough Conversations in Statistical Collaboration, taught by professors Julia Sharp and Emily Griffith, I thought it would be a useful topic for Statistically Speaking. It was! We had so many comments and questions for this episode, we couldn’t get to them all. So, our brilliant guests have agreed to answer more of them in this blog post.
Dr. Emily Griffith works in the Department of Statistics and the Data Science Academy at North Carolina State University and was elected to serve as the ASA Council of Chapters Governing Board District 5 Vice Chair beginning in 2023. Dr. Julia Sharp works in the Department of Statistics at Colorado State University and was recently selected as an ASA Fellow. She’s also currently serving as past chair of both the ASA Council of Chapters Governing Board and the Justice, Equity, Diversity, and Inclusion Outreach Group, which she formed in 2020.
Data is a human construct; you must identify the problem precisely for the best results. How do you ask the right questions to ensure the best results?
Emily: I don’t think there’s one right question or way to get the best results. I do think it’s important to build a relationship and to ask a lot of questions, show that you’re interested in the content, and allow for the people you work with to ask questions too.
Julia: I agree. Asking questions that help us understand the study objectives, design, and context is important to connecting with appropriate analyses. The “right” question might depend on several factors, including the relationship with the collaborator (i.e., is it a new relationship or a long-standing one), the power differential between you and the collaborator, the tone in which the question is asked, and how much we understand the context before asking the question. Because there are so many things that could contribute to asking the “right” question, asking many questions in an empathetic way is important to get the information we need to help guide and advise collaborators.
One issue we frequently encounter when gathering resource needs from data scientists/ modelers is eliciting sample size — for example, asking questions like, “How many samples do you need to address your hypothesis at the level of statistical confidence that you desire?” Often the research is so new that the scientist/modeler does not have a basis for confidently generating a sample size estimate. Any thoughts on how to move forward and align on sample size in this situation? In addition, large sample collection is usually time consuming and budget dependent; how can we convince customers to adopt a statistical approach?
Julia: Discussions about sample size can be tough! We have a video related to sample size in our collection of tough conversations training videos. In some circumstances, I have worked with enough researchers in a particular field to know that doing a sample size calculation will not be productive. For example, in working with some veterinarians, I know that in the past, there have been a maximum of six horses they can work with in one type of study. It would be counterproductive for me to do a sample size calculation to tell them they need 500 horses. My approach in these circumstances is to discuss effect size and what they can get from the results of their studies.
Emily: Great answer, Julia! I also use the effect size approach with my collaborators when the sample size is somewhat fixed due to cost, time, or space limitations. When this issue crops up, I would encourage you to consider asking more questions and really trying to understand what the cause is of resistance to a larger sample size. Sometimes things really aren’t feasible, but other times you can push a bit to get a larger sample size if you can demonstrate the value it would add.
On the issue of figuring out an effect size when a subject is new, sometimes I’ll just throw out a number: “What if the change was nine?” Then the subject matter expert will often reply with something like, “It would never be nine! I’d expect it to be between 0.1 and 0.8.” That technique has helped me nail down specific numbers by reassuring my own collaborators that while uncertainty is fine, they often actually do know what they expect to happen with a reasonable degree of precision.
When you listen and aren’t aggressive, what are ways to be perceived as not passive or a direct communicator?
Julia: Even when being an empathetic listener, we can show we are actively engaged in the conversation by providing nonverbal cues (e.g., head nodding), asking timely questions, and taking notes.
Emily: Absolutely. We can also wait our turn and then calmly and clearly state our opinions, thoughts, and findings. Direct communication doesn’t have to be aggressive, but it is important to be an active participant in your conversations to avoid seeming too passive or disinterested.
You mentioned keeping a running list of what you are working on. Do you have any guidance on which tools you've found to be most effective for such a list?
Emily: I use a streamlined bullet journal. My setup is basically a weekly to-do list, and then I add things as the week goes on and cross things out as I finish them. Each week I move any unfinished work over first, then continue to use the list throughout the week. That planning time on Monday is really helpful for me. I also put all my personal and work appointments on one Google calendar.
Julia: I’ve gotten out of the habit of using a to-do list, but I hope to reincorporate this into my routine this fall. I use my calendar to help with listing and blocking off time for projects. Like Emily, I use one calendar for both personal and work appointments.
We appreciate Emily and Julia taking the time to further the conversation through this blog post. They’ve even shared more useful resources below.
Though this material is primarily about collaboration around gaining insights from data, the takeaways are relevant for all of us. As George Bernard Shaw wrote, “The single biggest problem in communication is the illusion that it has taken place.” Julia and Emily’s wisdom will help you hone your communication and collaboration skills, and we think it’s so valuable that no registration is required! We hope you will watch this extremely useful episode and give us your feedback, as so many of the live participants did.
References and Suggested Readings:
ASA’s Ethical Guidelines for Statistical Practice
Derr, J. (2000), Statistical Consulting: A Guide to Effective Communication, Pacific Grove, CA: Brooks/Cole Press.
Sharp, J. L., Griffith, E., and Higgs, M. D. (2021). “Setting the stage: statistical collaboration videos for training the next generation of applied stat...,” Journal of Statistics and Data Science Education, 29(2): 165-170.
Sharp, J. L., Griffith, E., and Higgs, M. D. (2021). “Setting the stage: statistical collaboration training videos”
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.