George Rodriguez

Good advice is worth sharing when the objective is to improve the communities we work in, yet the manner in which feedback is given may determine if the advice affords the desired improvements or leads to polarization. The statistical community does a fantastic job of providing responsible guidance to those involved in data analysis of any kind. However, we should be vigilant of inadvertent minor condescension if we are to avoid alienating those we are trying to educate.

As a PhD chemist with a master’s in statistics and extensive experience applying machine learning in materials research, I often find myself in the middle of interdisciplinary discussions in which the wrong tone can undermine the value of such discourses. My immediate team consists of chemists, physicists, and data scientists, as well as chemical and mechanical engineers using a range of data science tools. The most important lesson I’ve learned from working with such a diverse group is the importance of respectful dialogue. I hope this commentary will help us be more thoughtful about how we give advice to ensure both sides benefit and lead to more inclusive dialogue between diverse and interdisciplinary researchers.

I have been participating in numerous webinars during the work-from-home movement precipitated by the COVID-19 pandemic. Many of these sessions have been given by experts from the traditional statistical community who provide guidance on proper data analysis necessary for avoiding common mistakes. Most pointers are motherhood and apple pie of statistical analysis and certainly worth discussing. Unfortunately, some of these tongue-in-cheek comments occasionally come across with a slight tone of condescension that can short circuit further discourse between those with good advice and those who may need it. For example, in at least three recent presentations, I’ve heard anecdotal evidence of humorous mistakes done by machine learning practitioners. Similar comments are regularly made about research involving statistical analyses by social and physical scientists. I tried to imagine how members from those communities felt as they listened to a talk they had attended to strengthen their statistical skills. It didn’t take me long to reach that level of understanding because I am also a member of those communities.

We all need to use our extensive experience in data analysis to guide the various communities using the statistical machinery developed throughout that long history. However, that guidance should be given with the same tone we want used when we receive guidance—the Golden Rule. Second, addressing the bad habits without calling out practitioners of any specific field allows us to target the problems while preventing academic polarization. Just imagine if statisticians were specifically ridiculed every time they found a beautiful pattern that was well known to those in the fields they’re supporting (i.e., manufacturing, biology, chemistry, economics, etc.). Such an approach would certainly derail potentially fruitful dialogue between people with different academic roots. These are the very same interactions we need to nurture the most in order to advance all academic communities using statistical methodologies.

I realize the irony of calling out a specific group (e.g., statisticians) while suggesting the avoidance of such call-out behavior. Yet, as generalists who work with researchers from many fields, statisticians and data scientists are in unique positions to set the correct tone for conversations between collaborators in highly interdisciplinary research. In this connection, I hope the approaches we use when giving advice about proper use of statistical machinery are as thoughtfully crafted as the advice given.