Syllabus

Course Times and Office Hours

What is This Course and Can / Should You Take It?

Innovations in statistical learning have created many engineering breakthroughs. From real time voice recognition to automatic categorization (and in some cases production) of news stories, machine learning is transforming the way we live our lives. These techniques are, at their heart, novel ways to work with data, and therefore they should have implications for social science. This course explores the intersection of statistical learning (or machine learning) and social science and aims to answer two primary questions about these new techniques:

1. How does statistical learning work and what kinds of statistical guarantees can be made about the performance of statistical-learning algorithms?

2. How can statistical learning be used to answer questions that interest social science researchers, such as testing theories or improving social policy?

In order to address these questions, we will cover so-called “standard” techniques such as supervised and unsupervised learning, statistical learning theory and nonparametric and Bayesian approaches. If it were up to me, this course would be titled “Statistical Learning for Social Scientists”—I believe this provides a more appropriate guide to the content of this course. And while this class will cover these novel statistical methodologies in some detail, it is not a substitute for the appropriate class in Computer Science or Statistics. Nor is this a class that teaches specific skills for the job market. Rather, this class will teach you to think about data analytics broadly. We will spend a great deal of time learning how to interpret the output of statistical learning algorithms and approaches, and will also spend a great deal of time on better understanding the basic ideas in statistical learning. This, of course, comes at some cost in terms of time spent on learning computational and/or programming skills.

Enrollment for credit in this course is simply not suitable for those unprepared in or uninterested in elementary statistical theory no matter the intensity of interest in machine learning or “Big Data”. Really.

You will be required to understand elementary mathematics in this course and should have at least some exposure to statistical theory. The class is front-loaded technically: early lectures are more mathematically oriented, while later lectures are more applied.

The topics covered in this course are listed later in this document. I will assign readings sparingly from Introduction to Statistical Learning, henceforth referred to as ISL. This text is available for free online and, for those who like physical books, can be purchased for about $25. Importantly, the lectures deviate a fair bit from the reading, and thus you will rely on your course notes much more than you might in other classes.

If—after you have read this document and preferably after attending the first lecture—you have any questions about whether this course is appropriate for you, please come talk to me.

What This Course is Not

The focus of this course is conceptual. The goal is to create a working understanding of when and how tools from computer science and statistics can be profitably applied to problems in social science. Though students will be required to apply some of these techniques themselves, this course is not…

…a replacement for EC420, EC422, or a course in causal inference.

As social scientists, we are most often concerned with causal inference in order to anaOpelyze and write policies. Statistical learning and the other methods we will discuss in this course are generally not well-suited to these problems, and while I’ll give a short overview of standard methods, this is only to build intuitions. Ultimately, this course has a different focus and you should still pursue standard methodological insights from your home departments.

…a course on the computational aspects of the underlying methods.

There are many important innovations that have made machine learning techniques computationally feasible. We will not discuss these, as there are computer science courses better equipped to cover them. When appropriate, we will discuss whether something is computable, and we will even give rough approximations of the amount of time required (e.g. P vs NP). But we will not discuss how optimizers work or best practices in programming.

…a primer on the nitty-gritty of how to use these tools or a way to pad your resume.

The mechanics of implementation, whether it be programming languages or learning to use APIs, will not be covered in any satisfying level of depth. Students will be expected to learn most of the programming skills on their own. Specifically, while there will be some material to remind you of basic R commands, this is not a good course for people who are simply looking to learn the mechanics of programming. This course is designed to get you to use both traditional analytics and, eventually, machine learning tools. We will do some review of basic programming, and you will have the opportunity to explore topics that interest you through a final project, but ultimately this is a course that largely focuses on the theoretical and practical aspects of statistical learning as applied to social science and not a class on programming.

Perhaps most importantly, this course is an attempt to push undergraduate education toward the frontiers in social science. Accordingly, please allow some messiness. Some topics may be underdeveloped for a given person’s passions, but given the wide variety of technical skills and overall interests, this is a near certainty. Both the challenge and opportunity of this area comes from the fact that there is no fully developed, wholly unifying framework. Our collective struggle—me from teaching, you from learning—will ultimately bear fruit.

Success in this Course

I promise, you are equipped to succeed in this course.

Learning R can be difficult at first. Like learning a new language—Spanish, French, or Chinese—it takes dedication and perseverance. Hadley Wickham (the chief data scientist at RStudio and the author of some amazing R packages you’ll be using like) ggplot2—made this wise observation:

It’s easy when you start out programming to get really frustrated and think, “Oh it’s me, I’m really stupid,” or, “I’m not made out to program.” But, that is absolutely not the case. Everyone gets frustrated. I still get frustrated occasionally when writing R code. It’s just a natural part of programming. So, it happens to everyone and gets less and less over time. Don’t blame yourself. Just take a break, do something fun, and then come back and try again later.

Even experienced programmers (like me) find themselves bashing their heads against seemingly intractable errors.¹ If you’re finding yourself bashing your head against a wall and not making progress, try the following. First, take a break. Sometimes you just need space to see an error. Next, talk to classmates. Finally, if you genuinely cannot see the solution, e-mail the TA. But, honestly, it’s probably just a typo.

Course materials

The course website can be found at https://ssc442Kirkpatrick.netlify.app (but you know that. You’re on it right now.)

All of the readings and software in this class are free. There are free online version of all the texts including Introduction to Statistical Learning (2nd Ed) and R / RStudio are free (don’t pay for RStudio). We will reference outside readings and there exist paper versions of some “books” but you won’t need to buy anything²

Evaluations and Grades

Your grade in this course will be based on attendance/participation, labs, weekly writings, and a final project.

The general breakdown will be approximately 55% for labs, participation, and weekly writings, and 45% for projects (see below for specific details). The primary focus of the course is a final project; this requires two “mini-projects” to ensure you’re making satisfactory progress. Assignment of numeric grades will follow the standard, where ties (e.g., 91.5%) are rounded to favor the student. Evaluations (read: grades) are designed not to deter anyone from taking this course who might otherwise be interested, but will be taken seriously.

Weekly writings are intended to be an easy way to get some points. Labs will be short homework assignments that require you to do something practical using a basic statistical language. Support will be provided for the R language only. You must have access to computing resources and the ability to program basic statistical analyses. As mentioned above, this course will not teach you how to program or how to write code in a specific language. If you are unprepared to do implement basic statistical coding, please take (or retake) PLS202. I highly encourage seeking coding advice from those who instruct computer science courses – it’s their job and they are better at it than I am. I’ll try to provide a good service, but I’m really not an expert in computer science.

More in-depth descriptions for all the assignments are on the assignments page. As the course progresses, the assignments themselves will be posted within that page.

To Recap:

Resources

Mental health concerns or stressful events may lead to diminished academic performance or reduce a student’s ability to participate in daily activities. Services are available to assist you with addressing these and other concerns you may be experiencing. You can learn more about the broad range of confidential mental health services available on campus via the Counseling & Psychiatric Services (CAPS) website at www.caps.msu.edu.

Accommodations

If you need a special accommodation for a disability, religious observance, or have any other concerns about your ability to perform well in this course, please contact me immediately so that we can discuss the issue and make appropriate arrangements. MSU has a specific policy for religious observance available here.

Michigan State University is committed to providing equal opportunity for participation in all programs, services and activities. Requests for accommodations by persons with disabilities may be made by contacting the Resource Center for Persons with Disabilities at 517-884-RCPD or on the web at rcpd.msu.edu. Once your eligibility for an accommodation has been determined, you will be issued a verified individual services accommodation (“VISA”) form. Please present this form to me at the start of the term and/or two weeks prior to the accommodation date (test, project, etc). Requests received after this date will be honored whenever possible.

COVID-19

Things are hard right now. You most likely know people who have lost their jobs, have tested positive for COVID-19, have been hospitalized, or perhaps have even died. You all have increased (or possibly decreased) work responsibilities and increased family care responsibilities. You might be caring for extra people right now, and you are likely facing uncertain job prospects.

I’m fully committed to making sure that you learn everything you were hoping to learn from this class! I will make whatever accommodations I can to help you finish your exercises, do well on your projects, and learn and understand the class material. Under ordinary conditions, I am flexible and lenient with grading and course expectations when students face difficult challenges. Under pandemic conditions, that flexibility and leniency is intensified.

If you feel like you’re behind or not understanding everything, do not suffer in silence! Please contact me. I’m available at e-mail.

Mandated Reporting

Essays, journals, and other materials submitted for this class are generally considered confidential pursuant to the University’s student record policies. However, students should be aware that University employees, including instructors, may not be able to maintain confidentiality when it conflicts with their responsibility to report certain issues to protect the health and safety of MSU community members and others. As the instructor, I must report the following information to other University offices (including the Department of Police and Public Safety) if you share it with me: • Suspected child abuse/neglect, even if this maltreatment happened when you were a child; • Allegations of sexual assault, relationship violence, stalking, or sexual harassment; and • Credible threats of harm to oneself or to others. These reports may trigger contact from a campus official who will want to talk with you about the incident that you have shared. In almost all cases, it will be your decision whether you wish to speak with that individual. If you would like to talk about these events in a more confidential setting, you are encouraged to make an appointment with the MSU Counseling and Psychiatric Services.

Acknowledgements

This syllabus and course structure was developed by Prof. Ben Bushong with iterative refinements by myself. All credit goes to Prof. Bushong. All errors are my own.

Miscellanea

All class material will be posted on https://ssc442kirkpatrick.netlify.app. D2L will be used sparingly for submission of weekly writings and assignments and distribution of grades.