Several times a year, students contact me to tell me that I was the worst professor ever.

To be precise, former biostatistics students contact me with the simplest, and often ignorant, statistics questions. These questions are so basic that it is clear that I have failed in my job as a stats professor.

With a basic dataset, a student might ask, “what test should I use?” Last month I had a student drop by my office with a result of p = 0.0071 to ask me to tell him whether or not his result is significant. Without a hint of irony.

If you taught comparative vertebrate anatomy, how would you feel if a recent student of yours came into your office and pointed at his biceps, and then asked, “What muscle is this?” This is how I feel when students come to me for statistical consulting.

My one-semester graduate biostatistics class doesn’t go into much depth and covers the standard, mostly univariate, frequentist statistical approaches that you find in similar courses. We spend a decent chunk of the course on probability, experimental design and why people do statistical tests and what the results mean. We spend a lot of time – and by a lot I mean a lot – on what p-values are, and the relationships among probability, null hypotheses, variance, distributions, and errors. (And when I say we spend time on it, I don’t mean that I lecture a lot about it. I mean students actively figure this stuff out. And their exams show that, at least at the time, they really understand it.) I am convinced that, at the end of the semester, that these students really understand the main concepts.

But then they don’t understand it after the course is over.

If a student came to me and said, “I realize that we didn’t learn how to do a GLM in class but from my reading I think that might be the best choice here, and I was wondering if you had the time to talk about it,” I’d ask her to pull up a chair. But when a student says, “I know I was in your class a couple years ago, but I’m looking at this dataset that I already collected and I don’t know where to start,” I’m not going to lift the paperwork that is probably occupying all three chairs in my office.

The hypothetical students-who-forgot-the-name-of-the-biceps-muscle are professional frustrations, and evidence of educational failure, for three reasons:

- They forgot something really simple. Though the name of muscle is a mere fact, is one thing you’d expect a student in a vertebrate anatomy course to remember, if not know before even starting the course.
- The students were intellectually lazy and didn’t decide to look up the answer, but instead just asked a former professor.
- The students demonstrated a personal lack of regard for the professor’s effort in teaching the course, by showing unawareness of the fact that the professor expects students to remember basic facts after the course ends and also empowered them to look up basic information. Or to put it in fake-biblical terms: the students had the temerity to think that we are there to feed them fish sandwiches instead of showing them how to fish and how to bake bread. In my view, this extends beyond personal laziness, by showing that the students don’t bother to show that they respect our role as teachers.

On the bright side, there is one positive aspect of the fact that students come in to ask me dumb stats questions. They think positively enough about my course that they think that I’m useful for statistical advice. That’s not much of an upside, but hey, it’s all I’ve got as far as I can see.

I’m long past taking this personally. I don’t get insulted when students come to me asking me to reteach a very basic concept of probability that we studied for a whole semester. I see that this is their problem, and not mine. I don’t take it home with me. There’s a 0% probability that this issue will keep me from sleeping. However, if I’m trying to be more effective at my job, then I need to confront the issue raised by these interactions. It’s a form of teaching assessment, in which I’m doing poorly. (Some students do thank me quite generously for what they learned in the course, and I’m not forgetting that input either.)

How do I handle these outrageous questions? It varies, because I haven’t (yet) developed an *a priori* approach to the situation. In some cases, I might simply chimp grin a bit and say something like “you really don’t recall how to test a null hypothesis?” or “So you’re telling me that you haven’t been able to find anything about what to do when your independent variables are categorical and your response is continuous?”

The bulk of the class is biomedically-oriented Master’s students who have some need for statistics with their thesis but don’t think that they’ll need to practice stats for everyday use. So, each semester, I make a point of saying at the outset, “When you design your thesis experiments, you’re welcome to consult with me about the process. But if you don’t discuss your stats with me before collecting your data, then there won’t be much I can do to help you.” I’ve had to remind a couple students of this fact, one of whom had a horribly pseudoreplicated design. And another who failed to run a necessary control.

While I’m not the most amazing professor, I don’t genuinely think I’m the worst, either. I just think that some students think it’s acceptable to offload course material from their brain as soon as the course is over, and do not feel obligated to go back to the hardware store if they lost the tools that they picked up during the course. And among students who are stats-phobic and math-phobic, which is a sizable fraction of the population in the course, they’re just glad they survived. It’s particularly frustrating because my guiding principle in this course is to teach a small number of fundamental concepts in a way that they are *supposed to* stick with the students for a long time. At least in this class, I know it isn’t happening, with at least some of them. I honestly don’t know what I can do, if anything, to make sure that students really remember what a null hypothesis looks like and what a p-value means. But it is clear that some of them genuinely forget it, presumably because they think it’s not important.

So, when I teach this class again in the fall, I have to find a way to make biostats personally important, with students who don’t see the usefulness of stats in their professional future. Wish me luck.

We get the same with our student too – great post

I felt similar this semester when, in an intro class (focused on genetics –> development —> evolution), I asked the students something along the lines of, “We see this pattern because each of our cells contain different DNA” and they resounded with a “Yes!” Facepalm.

Perhaps one strategy would be to tell them this story (i.e., that many of them will return to you in the future with questions) on the very first day of class. Tell them that stats are very much a “you don’t use it, you lose it” discipline, AND THUS (this is where you wring your hands emphatically at them) the best way they can help their future selves is to take REALLY good notes. Tell them to annotate their own assignments (and/or code, if you’re using it) with explanations much more basic than they think they need, because a year from now it will be at exactly the right level for them. Then tell them if they want to meet with you in the future, you’ll ask them to bring that notebook with them (at which point you can have them sit and look the answers up themselves in their own notes while you continue with whatever you were doing). That way you are there to help interpret if necessary, but hopefully it discourages the people wanting a fish sandwich.

My grad stats professor (shoutout to Andrew McAdam of U Guelph) gave us this advice (to copiously annotate), and many of us who took it are still using our notebook from that class, years later. (After seeing how much I relied on it, I even later went back and indexed the whole thing so I could quickly flip to things: “random effects,” “standardizing,” etc.) Anyway, maybe the key is a preemptive strike? Just an idea.

Good luck! 😉

It’s not you, Terry. The students are failing at “far transfer,” which means they’re cramming for exams, memorizing formulas and process, but “flushing their buffers” (so to speak) when the semester ends and they’ve “completed” the course. See Susan Ambrose et.al.,

How Learning Works: Seven Research-Based Principles for Smart Teaching.When I was an undergraduate, I was somehow* invited to participate in a “Task Force on the Undergraduate Experience,” where a bunch of people gathered in a circle to talk about things like the progression from large survey courses up through senior thesis-writing. I know the effort involved getting faculty together to talk about knowledge transfer across courses, but I

alsoknow that not all institutions have forums in place to have these kinds of discussions. Tufts was actually pretty good at it.It would make me want to pull my hair out in grad school, when I would talk to seniors who took introductory lab courses with me, and who wouldn’t remember the basic structure for writing a lab report or the equivalent paper, despite the fact that in the introductory series, we spent time reading and deconstructing REAL papers written by REAL scientists, and practiced writing multiple reports in the same format (which the students mastered by the end of the course). But then, I switched from introductory courses to intermediate courses, where the bulk of the coursework was problem-solving, without requiring any experimental design or writing skills, and I see that that’s the stage where the students lose track of these things. Why aren’t the Genetics students still writing? Why aren’t they still doing statistics while they’re solving Genetics questions (we make them start doing that in the intro courses)? Is it that important that we shove in that much other information in Genetics, that the other skills fall by the wayside? I guess my concept of what constitutes a well-trained scientist differs from the concept of my colleagues!

If we as instructors aren’t working to make connections with each other across the curriculum, is it really a surprise that students aren’t making those connections, either?

*I’m an opinionated person. 😉

The reviews for that book looked good, so I just ordered it.

Oh, but I structure pretty much the entire course with the goal of long-term retention. They don’t cram for exams because the final is a take-home, and they have a full list of possible questions before the other exams. There is scant requirement for memorizing process, and they don’t have to memorize anything because they typically get to take notes into the exam. I haven’t read Ambrose et al yet, but I am doing (or earnestly trying to) everything to make sure that the class isn’t focused on the short-term and the gamesmanship of earning a grade. That’s why this is particularly frustrating.

Me too.

In my opinion the overall problem lies not with a single course or teacher but in the way college students approach learning in general; as something to get through with minimum effort and time so that they can then return to the rest of their lives.

That said, an achievable (partial) solution for your problem might be to make a slide with a diagram on how to choose the right test. After your course your students should be able to understand the information they find online or in their notes, but may not know where to start looking for their specific dataset. Something as simple as a stats key might be helpful. Think of it as outsourcing the questions you ask them when they show up in your office; ‘Is your data continuous?’, ‘How many independent variables do you have?’ etc. After this, they can google ‘anova’ themselves and read what they need about normalcy, variance etc. It will not help those students that show up with out knowing what significance is (or those that are too lazy to open their laptops), but it may help some of the others be more independent and critical in their approach of stats.

I think this is common for math classes. Several years after taking a stats course I was like your students. Once I started analyzing my own data it became real, and sunk in more successfully. Even screwing up, realizing how you screwed up, and having to start from scratch, will lead to vast improvement in retention. I doubt it has anything to do with your teaching, and I bet they ‘learn’ a lot faster the second time around. The only good solution isn’t practical: teaching the course concurrently with the initial planning stages of each student’s thesis, then a review after all the data collection, as they begin their analyses. Leslie’s solution sounds promising though and certainly more practical.

Terry, I took your biostats class 3 years ago. I agree with Leslie. After I had my data, I spent a great deal of time looking back at my class notes trying to figure it all out. You teach the class as a seminar, which I’m sure is great for many learning styles, unfortunately it wasn’t great for me. I just had a hard time (read as, “I had to study my ass off for that A”).

Don’t get me wrong, though. You are a great teacher, who is excited about the subject and has fantastic real world problems to solve. I don’t know what you could do differently.

As EcoGrad commented, it did sink in more when I had to analyzed my own data, and even more when I ended up being the go to person for other grad students. Let’s face it, stats is not easy or intuitive for the average student.

Great post! I agree and feel we have quite similiar problems. To me its strongly related to the way statistics and methods in general are implemented into the curriculum. In Germany, virtually everybody hates the basic statistics courses. A lot of theoretical munbling, but they do not learn the real deal. If we do not derive curricula where they learn to set up proper designs, analyse the data and write about it, we fail – big time! Current curricula are not made to tackle this. I say we need big changes, throughout disciplines. However right now its only smaller and more disciplinary courses that clean up some of the mess. I dream of a really hands on first semester stats course that teaches what every student should need: Hypotheses-driven reproducible statistics. Hope we achieve our goals!

This is sort of a separate problem, but it’s related, and I could really use some feedback. I’m a teaching postdoc at a big state university, doing half basic behavioral ecology research and half undergrad course development. I felt like stats was one of the aspects of doing a PhD I was least prepared for, even though I took 3 different stats classes in college. Luckily, one of them was a truly fantastic experimental design course, however, which at least helped me set up my experiments correctly, even though I forgot a lot of the details about particular tests and had to relearn them again once I actually had data.

I’ve been trying to incorporate that kind of stats instruction into the classes I’m helping develop (i.e., “this is what a confounding factor is, and this is what a p-value means,” not “this is how to do a t-test by hand”). I thought the idea that biology undergrads need better grounding in statistics would be uncontroversial, but I’ve run into a lot of resistance from some of my colleagues, most of whom are from biomedical or molecular biology backgrounds (most bio majors at my institution are pre-med or interested in biomedical research). Other postdocs in my program have gotten the definition of a p-value wrong, and I had to explain the concept of a paired t-test to a colleague whom I know to be a real whiz in the lab, when she was trying to analyze an educational intervention she’d done. One of the directors of the program reprimanded me for focusing so much on stats, because she doesn’t think it’s a skill students will need to use in the future. Obviously, these people have all been successful as biologists, despite being, from my perspective, surprisingly ignorant about statistics.

It’s been bothering me, and it makes me question myself as a teacher. On the one hand, I want to prepare students for the careers they’re seeking, and if statistics really aren’t as important for biomedical folks as they are in EEB (where everyone I’ve talked to agrees with me), maybe I should back off. On the other hand, my strong instinct is that learning statistics helps scientists (and probably non-scientists, too) reason carefully and interpret the results of studies accurately, so I should push for it to be included in the curriculum.

Sorry for the super-long comment – this post just came a a perfect time for me, because I was actively thinking about the issue. Does anyone have a perspective on whether ecologists and evolutionary biologists are just unusually nuts about stats? I could imagine that being true, because field experiments are often unbalanced and include many uncontrolled factors, so they can be challenging to analyze. Do students in other biology disciplines need to learn statistical reasoning as badly as EEB students do, or am I just biased by my own experience? Thanks, and thanks to Dr. McGlynn for the well timed post!

Thanks so much for this book recommendation. I already am partway through it. It’s crazy good, and really practical, and applies the science in

How People Learnto things that can be done in the classroom.