The stats guy

References

Cite

Share

Abstract

J and Carlos, two ecology professors at Enormous State University, are passing the line of graduate students waiting to see Kai, the statistician. “Ah, spring”, says Jane, “when a young grad student’s fancy turns to ANOVAs and Bonferroni corrections”. Like many institutions, ESU’s ecology program regularly advises students to consult with a statistician to help with experimental design and data interpretation. Of course, the students also have to take a couple of stats courses, but most admit that at the end of they still feel pretty shaky about statistics. As a result, nearly all of the students in ESU’s ecology program rely on direct advice from Kai in designing their studies. Although most of the ecology faculty have some facility with statistics, Kai is recognized as the local statistics guru. His role in the department has evolved over time, from occasional consultant to a fixed part of the educational process. In fact, most graduate committees in the ecology program won’t approve a student’s research proposal without the coveted “Kai stamp of approval”. Jane and Carlos had often disagreed about the role Kai plays in graduate training at ESU. Jane was more critical. “I think our students rely on Kai for too much. If you can’t understand enough statistics to interpret the data from your own experiments, then you probably don’t deserve a PhD in ecology. Besides, we are setting a truly dangerous precedent for these students – that someone else can be held responsible for your results. Carlos, you were at that student seminar last week; he couldn’t answer questions about his design beyond the basic level. He couldn’t even begin to answer my question about why the factors in his analysis were fixed versus random. All he could say was that Kai told him how to interpret his output and what all the stats meant! Our students publish work whose fundamental statistical design they don’t understand and can’t defend. If we’re going to train grad students, we have an obligation to train them to be scholars. What’s more important than that, though, is that they need to understand that they must be able to defend every aspect of the research that they publish.” Carlos shrugged. “So, what are you proposing? Should students take 2 years of stats courses? How about 3 or 4 years, just to be sure they get it? Be reasonable. We’re an ecology department, not a statistics department. Ecology has become complex, computationally and analytically. There’s nothing wrong with a statistician becoming part of the research team. Every member of the group has a role to play. Part of the value of collaboration is that everyone can’t be expert in everything. Our PhDs should understand the big picture, but I’m not too worried if they don’t become statisticians in our program.”