The case method and statistics

Readers of my books will not be surprised to hear that I am a fan of the case method in teaching. The case method was pioneered by Harvard Law School in 1870, and I was exposed to it at Harvard Business School, where it is used to teach everything from marketing to leadership, and from economics to operations, and even accounting.

Since the 1920s, the case method replaced the "lecture and drill" style of teaching at HBS. It is a radical approach which requires students to think deeply about the case materials, engage one another in constructive debate, and draw general knowledge out of specific examples without getting drowned in the minutiae. It is radical in that the professors do not present any slides or lectures at all.

Let me describe a typical class. Case materials are distributed in advance of the class. Students are responsible for reading the case, and preparing detailed responses to some open-ended questions based on the case. The most important question is always "what would you do if you are faced with this business decision?" The case materials are presented as a narrative, often divided into a sequence of multiple stories, each one ending with a key moment in the company's history. The goal is to simulate the decision-making environment of the protagonist of the case. This means irrelevant data is included. A proper response starts with an "action plan" describing the student's decision, and what follows is the argument marshalling the facts of the case, any analysis of the data, and attention to specific details of the company, the culture, or the environment.

The teacher enters the classroom of 70 to 90 students, and alerts one that he or she will lead off the discussion. The teacher acts primarily as traffic cop during the class, encouraging students to debate each other, occasionally nudging the direction of the dialogue, but takes care not to present a "correct" answer. The teacher uses the blackboard to organize the key ideas as they are raised by students. If a key idea never surfaces, it will remain untouched; it is considered bad form for the teacher to stop the flow of conversation and introduce a foreign concept. A minority of teachers reserve the last five minutes for a summary--but the summary is almost always of the issues, the conflicts, rather than a clean description of the "correct" course of action. Sometimes, it is just a review of what the protagonist in the case actually did, and whether the actions taken panned out.

***

The case method is controversial. Students who hate it typically complain that they "learned nothing." They are looking for a bullet-point summary of the day's learning. They believe in the black-and-white view of the world: there is a "right" way to do something. They show little patience for uncertainty.

Others complain about the inefficiency of the case method. As you can imagine, there are long stretches of time in which the class is collectively confused.

***

I like the case method, and I think it is particularly suited to teaching statistics.

One of the key philosophies of the case-based teaching is that the business environment is filled with uncertainty, with incomplete knowledge, and therefore there is no single right way of doing things. That sounds like a typical statistical problem!

The case method also acknowledges that general principles are useless in running a business unless they are adapted to specific situations. Business managers must adapt to changing circumstances and cope with uncertainty.

***

In writing my two books, I have taken half of the case approach--the use of case studies, but also adapted it to the printed form. The Socratic method is missing, and I've included a few takeaways. I believe strongly that you can't reducing managing a business to a cookbook, nor can you provide detailed recipes to solve business analytics problems.

The case method demands a lot from users, and that's why 50% of grades at HBS are determined by class participation. You'll get the most out of Numbersense if you engage with the materials actively, discuss them with other readers, figure out other ways to interpret the data, and look for general principles.

***
A good article about how several Harvard professional schools implemented the case method is here. What strikes me are the different ways of expressing the same philosophy. Forbes has a nice piece last year with an inappropriate title.

 

Comments

[Glenn]

I am interested in hearing more about this in a stats class. Besides your book, are there any other resources you could recommend? I can see lessons in this style going over very well in an AP stats class in high school.

[Kaiser]

Hi Glenn, I'm hoping that others might chime in. There are few useful materials right now because we are far from recognizing the liberal arts component of statistics. Many textbooks contain "case studies" which are no more than long examples stripped of real-world complications, not in the style of HBS cases.
If you read the article I linked to at the end, it gives a history of how the case method got adopted at Harvard, and one of the prerequisites is developing a body of suitable cases. I suspect that business cases will be hard to come by because businesses don't want to publish their data. (Also, it is hard to fake multidimensional data.) I suspect that it may be easier to develop cases from nonprofits or other arenas.

[Rclickhandbuch.wordpress.com]

I think this is a very good auxiliary method. But it can't replace some "classical" teaching.
Here's an example: I first need to explain the concept of linear regression, before students can use it in case method style.
This teaching could include the mathematics behind the parameter fitting, the R-commands necessary (my students don't know "~" for formulas before they learn lm), the formula behind Rsquared and it's interpretation, what a correlation is anyways, ...
I doubt students should learn this on their own - teaching time is too valuable for that. Let's get to the application of statistical methods!

[Kaiser]

Rwc: We have to distinguish between teaching data analysts and teaching literacy. I think the case method works for both but for the former set, classical teaching is needed as well.

The case method as practiced at Harvard relies heavily on group learning. What I didn't cover is the study group. Four or five students get together every morning to discuss the cases before they go to class. Among the study group, it is usually wise to have both analytical types and qualitative types so the knowledge is shared.

Such group learning is easier for professional schools in which students have a wide variety of backgrounds. For say college courses, I'd recommend including handouts and exercises that students should work on before class but using the class time for case discussion rather than details of formulas.

I'd suggest that an intro stats class along these lines would be much more engaging and effective than the standard one.