Economics professor Shoshana Vasserman is one of three faculty members who teach Data and Decisions. | Elena Zhukova
Fear not, Lanier Benkard told his students halfway through the winter quarter, they would learn regression analysis even if at that moment it seemed like an impenetrable black box. “You probably feel ill-equipped to do this right now, because you are,” said Benkard, a professor of economics and the designer of the base MBA1 course Data and Decisions. “But it will start to make sense soon.”
Editor’s Note
In this ongoing series, we bring you inside the classroom to experience a memorable Stanford GSB course.
In an age in which data analysis is crucial for almost every industry, Benkard’s course aims to give students a firm grasp of the tools and approaches they will need to assess performance and build predictive models to inform decision-making.
“Fifteen or twenty years ago when we tried to teach this stuff, people thought, ‘I don’t need to know how to do that. I will hire somebody else to do that.’ I don’t think anybody believes that today,” Benkard says. “You can’t say that you understand a machine learning model if you have never run one yourself.”
Lanier Benkard redesigned the course five years ago. | Nancy Rothstein
All first-year students take some version of Data and Decisions. Roughly three-fourths enroll in the base course, while the rest opt for the accelerated or advanced sections. Benkard redesigned the base course five years ago as a “flipped” classroom — students absorb most of the material in online modules, including videos that describe in detail everything from how to use R software to the intricacies of hypothesis testing. Class time is devoted to answering questions and giving students plenty of practice working with data sets.
In one class midway through the quarter, Benkard used examples from Moneyball — the data wizardry that revolutionized baseball, popularized in the Michael Lewis book of the same name. It was not important that students knew the game — indeed, at one point one asked what “on-base percentage” referred to — only that they understood why and how the Oakland A’s management found a clear, compelling relationship between how often batters reached base and the number of runs the team scored.
Benkard cold-called half a dozen students at various points to test their understanding of, for example, what was required to build a model with a 95% confidence level. In other class periods, students examined the relationship between CEO compensation and stock performance, how to predict house prices, and what factors affect movie ratings. Students were given 45 minutes to load the data, conduct regression analyses, and come up with answers. “Every day, they are working with a different data set and a different set of questions,” Benkard says. “By the time the class is finished, they’ve done this 30 times or so, and it’s empowering because now they know how to do this.”
“I had limited experience with regression analyses before the course,” says Walter Winslow, MBA ’24. “Re-learning them in a business context has been super valuable to me. It has made me much more well-rounded in data and analytics and a much better future team member who can apply the learnings in real time.”
Fully half of the course is devoted to regression analysis. “I think of regression analysis as what we want to teach and the rest of the stuff as getting us there,” Benkard says. “There are a lot of fancy names in machine learning that sound sexy, but they’re really just different regression models.”
By the end of the course, students will have a new suitcase of skills and a richer vocabulary — coefficients and scatterplots and confidence intervals. The black box slowly sheds its opacity. “We want students to understand what’s possible with big data, but also what the limitations are, how to do it intelligently, how to recognize when somebody is doing it badly,” Benkard says. “It may look like magic, but it’s not.”
For media inquiries, visit the Newsroom.
