Brittney Bailey is a biostatistics PhD student at The Ohio State University who is interested in the analysis of clinical trials with partially clustered designs. As a graduate student researcher, she collaborates with researchers in the Stress & Health Lab of the Wexner Medical Center.
As students, we anticipate differences between practicing statistics in the classroom and practicing statistics in the real world. We know we will no longer have our textbooks or professors providing us with clearly phrased questions that we are prepared to answer. We know the data will not be as perfect or as small as those textbook examples. We know we will work with nonstatisticians, some of whom think statistics is magic. But anticipating these differences is not the same as experiencing them—the reality can be overwhelming when we start practicing statistics outside the classroom. Come along with me as I provide tips to ease the transition from student to collaborator based on my own experience in a collaborative research lab.
Learn to communicate.
In a collaborative setting, you are often working with nonstatisticians to help answer their questions. You may be brought onto a project at any time, from developing hypotheses and designing a study to running analyses and interpreting results. Regardless of when you are brought on, the first step is discussing the project with someone else. If the field is unfamiliar to you, you should learn enough about it to be able to communicate with your collaborators. Read prior literature or study protocols produced by your collaborators. Google terms or ask your collaborators to explain concepts you do not understand. There may be a lot of back-and-forth as both you and your collaborators become familiar with different terms or concepts.
You are also responsible for interpreting results and making sure your collaborators understand both the results and, to a certain degree, the methods used to obtain the results. You may need to explain the methods or results in several ways, whether it’s by using graphics and tables, modifying your wording until they get a better understanding, or tying more abstract ideas to real-world examples.
Establish good working relationships.
Good communication must be maintained for the duration of the project. Be clear about what you expect from your collaborators and what they can expect from you. Take the time to clarify anything you are unsure about and rephrase ideas to ensure you are all on the same page. Respond to messages in a timely manner, even if it’s just to acknowledge receipt of the message, and provide progress updates if a task takes longer than expected. Be prepared for meetings by reviewing the project beforehand and having materials ready to present.
Prepare for difficult conversations.
In some cases, you may have to deliver bad news. Maybe the study is not feasible due to sample size limitations or the data are not salvageable due to missingness, measurement errors, or poor data collection. Maybe you made a mistake at some point in your analyses and the results are no longer significant. In other situations, you may not be comfortable with a task due to ethical or scientific concerns. These are not easy conversations to have, especially when collaborators are under pressure to produce results.
Confidence in your own ability will help you be more productive and alleviate the stress that tends to accompany a lack of confidence.
Try to have difficult conversations in person when possible, and maintain a respectful and professional tone in all forms of communication. If you make a mistake, own up to it quickly and make every effort to resolve the problem. When it comes to ethical or scientific concerns, be prepared to stand your ground. You may have to defend your position, but you should not do anything you are not comfortable with. If you are not sure whether your concerns are valid, discuss them with a trusted colleague or mentor before bringing them to your collaborators.
Be patient with less-than-perfect data.
Real-world data can be awful. You may spend a significant amount of time restructuring the data file, renaming variables, and cleaning the data just to get it in a usable format. Before doing anything, be sure to save the original data to a folder where it will never be edited.
Even with an excellent data management system, real-world data are subject to missingness, limits of detection, outliers, and other errors that make analyses more difficult (or sometimes impossible). In many cases, the data do not satisfy the assumptions needed to perform your planned analyses. You might be tempted to transform the data or try a more complex method of analysis, but you must keep in mind the goals of the collaborators and the limitations of what they are comfortable with you doing. You will need to be able to explain your choices to your collaborators, and they will also need to be able to present their results to others. Your approach will often be a balance between correctness and simplicity.
Continue to learn.
Remember that learning never ends. If you have the benefit of working with a mentor when you begin, take advantage of the opportunity to learn from them and get feedback on your performance. If you are not sure about a method, take the time to read about it—review old notes or textbooks, research online, or reach out to a mentor or colleague. Stay updated on current methods by occasionally browsing journals or conference proceedings. Attend conferences, workshops, and short courses when possible—these are great opportunities to get new ideas and connect with other statisticians and potential collaborators.
Be confident.
Confidence goes a long way. Confidence in your own ability will help you be more productive and alleviate the stress that tends to accompany a lack of confidence. If you interact with your collaborators and present your work with confidence, your collaborators will have more respect for you and more trust in your work. It may take time to build confidence, but you should trust that the years you spent training in statistics have prepared you well for your collaborative role.
I am completing my third year of PhD program in Statistics at the Penn State University, and I do second Brittney in her thoughtful comments. In fact, at Penn State, as part of Statistics PhD program requirement, the students indeed go through Statistical Consulting training for two semesters. In the second half of that training, the students play the role of Statistical Consultants at our university’s Statistical Consulting Center, where researchers across Penn State Campus (and on some rare occasions from outside the Campus) do visit them for various sorts of statistical advice regrading their research studies. Based on my experience in that, the points Brittney mentioned do indeed come out to be super helpful.
Thanks again Brittney!
Very interesting and great piece!
Great blog, a really interesting read. Lots of this advice suits statisticians of all ages and experience – not just students!!