course
Data Analysis

Dec 2013

Professor Eirich takes pride in his work and cares deeply about teaching. It is obvious that he puts a huge amount of thought into the course. What's more, he is enormously friendly and an empathetic, engaging lecturer. The course is mostly a survey of regression techniques at either a high introductory or low intermediate level. The material is tailored to social science research and the assignments are designed to facilitate an outside research project like a masters thesis. Professor Eirich is a savvy statistician and a dedicated sociologist, and exposure to high-quality statistical practice is worth at least as much as proofs and matrix notation. He is also willing and able to answer both highly technical and highly intuitive questions, and will gladly go out of his way to find answers that he can't produce in class. Professor Eirich and his Data Analysis course are a huge asset to the QMSS program that I would recommend to anyone looking for an introduction to linear and logistic regression.

Mar 2010

This course is tragic. It is suppose to be some sort of applied regression analysis class, but has a massive identity crisis that leaves everyone at a loss. Eric Johnson lacks any coherent pedagogy other than reading long paragraphs from a PDF. He seldom motivates material, and when he does he utilizes some strange religious congregational data set that often abstracts from the points he's trying to get across--as if they're discernible in the first place. Most importantly, given the ample confusion in class he passive aggressively resents students who ask questions or get confused by the presentation of material. The end result is a class that is always lost, but also a classroom that is poisonous. The nitty gritty of the material: the lectures are incoherent. Eric does not quite know who he lecturing too. He is far too verbose with trivial points and finalities, yet takes large mathematical and theoretical concepts as given--they appear out of thin error, confuse everyone, and then disappear just as quickly. Unlike most econometrics or regression analysis courses, equations do not build on one another or refer back to previous statistical building blocks. The result is disjointed bits of information that are never really tied together. For instance, we're given the most rudimentary explanation of heteroscedacity, without real motivation and with equations that have appeared out of thin air. Concepts come out of left field, before vaporizing into a wall of meager mumbling. Information is so decontextualized and disjointed it even makes self study hard. It is as if someone has produced lecture notes by throwing darts at a regression textbook, or a REAL regression class syllabus. Go home. crack open any regression or metrics text, and you realize how little you learn and how bastardized the material is. I've had bad classes, but this is king of shit mountain. Given it's part of a core curriculum and a "keystone" class, it adds a nice sting. The jokes on you. *This class, however, is a nice lesson in watching an entire graduate class solve the collective action problem of revolting against a professor.