Introduction 3

What is statistics? Why should I study statistics? 4

Mathematics 5
Computer Science 8
Statistics 10
The Concentration 16
Planning Your Statistics Courses 17
Sample Course Plans 20
Concurrent Master’s (A.B./A.M.) 22
Research 23
Applied Research 23
Theoretical Research 24
Finding An Advisor 25
Funding and Getting Course Credit 26
Contacts 29
Advice from Current and Former Students 30
Acknowledgements 32
Introduction
Hello person-who-might-be-interested-in-statistics!

The Group for Undergraduates in Statistics at Harvard (GUSH) has

created this short (uh…it’s gotten a lot longer) booklet as a way to
introduce students to the wonderful world of statistics at Harvard. We
understand that students who might be interested in statistics courses
come from all kinds of math/statistics backgrounds and with all kinds of
goals and concentration interests, and we hope that this guide will help
you understand what statistics looks like at Harvard and how you can
explore and get started with different courses, research opportunities, and
concentration options.

While we’ll try to be as comprehensive as possible, this guide is by no

means meant to serve as a replacement for other advising support like
talking to concentration advisors, professors, and fellow students. That
being said, feel free to reach out to us with any questions you might have –
we’re here to help make your experience with statistics at Harvard as
enjoyable as possible!

With lots of love,

Harvard GUSH
harvardgush@gmail.com
gushclub.org
What is statistics? Why should I study statistics?
Ask different people, and you’ll probably get a wide range of definitions.
At Harvard, the statistics department will expose you to various main
ideas: probability, where you’ll learn to reason about and “control”
random events, numbers, and objects; statistical inference, where you’ll
apply your knowledge about probability to draw inferences from data; and
a variety of other areas that use some combination of the two (e.g.,
machine learning, biostatistics, data science). Harvard’s own website notes
that “the Department of Statistics helps students acquire the conceptual,
computational, and mathematical tools for quantifying uncertainty and
making sense of complex data arising from many applications.”

Especially as the influence of data and its analysis grows, statistics has
become more important than ever. Learning statistics will give you better
insight into how “statistical analyses” come to be: through the
concentration, you can learn what a p-value is, what confidence intervals in
political polls capture, how to compute how much you’ll win from a
gambling game, what linear regression is really “doing,” and much more.

From our own experience (and this can easily be confirmed by most people
at Harvard), the things you’ll learn in the statistics department are super
useful – and often necessary – for CS classes (e.g., CS 121, 124, 181, etc.),
for econ classes (e.g., Ec 1011, 1123, 1126), for internships/interviews/jobs
(e.g., in finance, SWE, biotech, data science), and for so many other things
that we simply don’t have room to list here.

Besides this, we’ve found – and we hope you will too – that statistics is a
beautiful way of using math to learn about the world and other subjects.
You’ll also find that learning other subjects makes you a better statistician.
Courses in Statistics and Related Fields

Mathematics
Math forms the backbone of statistics, and a solid foundation in linear
algebra and calculus is extremely helpful and almost necessary for most
statistics courses at the level of Stat 110 and beyond.

Math Ma/Mb: Introduction to Functions and Calculus

Math Ma/Mb are designed for students who aren’t ready for Math 1a but
want/need to learn calculus. The two course sequence – which roughly
covers precalculus and calculus – first teaches the mathematical
fundamentals for calculus and then covers the content of Math 1a. After
taking Math Ma/Mb, students are ready to take Math 1b.

Math 1a: Introduction to Calculus

Math 1a is the introductory calculus class at Harvard and is the first half of
a one year sequence in single-variable calculus. In 1a, differential calculus is
the focus. The class begins from elementary mathematical modeling,
highlights the central role of rates of change to motivate the concept of
derivatives, and culminates in the fundamental theorem of calculus and
applications (including integrals and integral rules).

Given that differential calculus is a prerequisite for Stat 110, if you have
never seen this material (or are very rusty), you should take this class. In
particular, if you didn’t take AP Calculus BC, then you probably will have
to take this class. If you took it but are rusty, then refer to your Math
Placement test to see if you still need to take it. In any case, before taking
Stat 110, it would be helpful to at least review derivative rules and the
fundamental theorem of calculus, as well as integral rules.

Math 1b: Integration, Series and Differential Equations

The second half of the first-year single-variable calculus splits time between
integral calculus, series, and differential equations. The first two thirds of
the course contain material that is relevant for Stat 110, such as Taylor
series, convergence / divergence tests for series, and integrals. Even so, a lot
of the material on integrals relevant for STAT 110 is covered in Math 1a.

While a proactive student could reasonably go from Math 1a to STAT 110

by self-studying the aforementioned relevant material, it can still be smart
to take Math 1b before taking Stat 110 if you have little to no familiarity
with series and integration. Any questions regarding mathematical
placement should be referred to the Math department. It is additionally
highly recommended that you refer to your Math Placement test results;
these will give you a gauge as to where you should begin.

Math 21a: Multivariable Calculus

Math 21a is an applied introduction to multivariable calculus. You’ll spend
a lot of time extending the knowledge of single-variable calculus you’ve
built to multiple dimensions. For instance, we might be interested in
taking a partial derivative of a function of multiple variables, which
captures how the value of a function changes if we change only one input
infinitesimally and hold the other inputs constant. You’ll also learn to take
integrals over more complex surfaces.
Math 21b: Linear Algebra and Differential Equations
Math 21b is an applied introduction to linear algebra. As opposed to
22a/25a/55a, Math 21b focuses primarily on the computational aspects of
linear algebra. You’ll learn how to work with linear transformations of
multidimensional space, manipulate matrices meaningfully, and how to
connect matrix operations to differential equations.

Math 22: Vector Calculus and Linear Algebra

If you’re interested in dipping your toes into theoretical math but aren’t
too keen on having most of your free time consumed by actually doing
math, then you might want to consider Math 22a (linear algebra) and
Math 22b (multivariable calculus). The class assumes no prior knowledge
in proof-writing and starts from the basics, providing a solid foundation
for anyone who is interested in learning how to write proofs. Most
students have not previously taken linear algebra or multivariable calculus,
although students who did have prior experience with these topics in a
non-proof context did not find the Math 22 material repetitive.

Math 25: Theoretical Linear Algebra and Real Analysis

A fast-paced and rigorous introduction to linear algebra in the fall (Math
25a) and to real analysis in the spring (Math 25b). Math 25 problem sets
will eat up a decent chunk of your free time, but you’ll leave the class with
a solid and theoretical understanding of linear algebra and real analysis,
and you’ll feel confident writing proofs. The only truly necessary
prerequisite is single-variable calculus at the level of AP Calculus BC, and
the course does not expect students to have any prior experience with
proofs. Some students come in having taken courses at the level of 21a/b
and having done some competition math, but many students come in
having done only one or none.

Math 55: Studies in Algebra/Group Theory/Analysis

Math 55a covers linear algebra, group theory, and some basic
representation theory whereas Math 55b covers topology as well as some
real and complex analysis. Both semesters are fast-paced and very rigorous,
so be prepared to spend a lot of time on problem sets. The midterm is
typically given before the add-drop deadline, which makes it a good
diagnostic for whether this class is right for you. There are no formal
prerequisites, but experience in writing proofs would be helpful. While all
topics in this class are foundational to anyone wanting to pursue further
study in pure mathematics, statistics majors will typically only encounter a
handful of the topics in their coursework.

Computer Science

CS 50: Introduction to Computer Science

Ah yes, the classic Harvard CS 50 experience. Taught by the incredibly
theatrical David Malan, CS 50 has one of the highest course enrollments at
Harvard. This course is geared towards beginner programmers and covers
many languages (C, Python, SQL, HTML, CSS, JavaScript) and
introductory topics (data structures, algorithms, memory, databases, web
development) in one semester. Many would say CS 50 is too fast-paced for
a complete beginner who has never touched CS before, but if you dedicate
time and effort to working on the labs then it’s a fruitful experience. If you
have taken AP Computer Science in the past then the topics might be too
basic, but be sure to check the websites of past CS 50 iterations to evaluate
whether the course content could be beneficial for you.

CS 181: Machine Learning

You’ve probably heard the term “machine learning” countless times, but
what actually is machine learning? CS 181 gives a fundamental (albeit
perhaps sometimes unglamorous) introduction to what machine learning
is. You’ll learn about the foundations and mathematics of machine
learning, starting from basic clustering and regression methods, which
then give rise to more flashy ideas like neural networks. While you might
not learn all of the state-of-the-art ML ideas (e.g., CS 181 usually doesn’t
cover transformers in depth), the course gives you enough background to
pick these up. Being familiar with manipulating expectations and
probabilities is a must, as is working with matrices and vectors. Likelihood
ratios figure fairly largely throughout the course, so having taken or
concurrently taking Stat 111 can be helpful (and vice versa), though it
certainly is not necessary.

CS/Stat 184: Introduction to Reinforcement Learning

Are you curious about how computers beat humans at Go, chess, and
Starcraft? Have you wondered what gives a language model chatbot its
"personality"? Reinforcement learning (RL) lies at the heart of these
achievements. This subfield of machine learning investigates how an agent
can learn to make decisions in an interactive environment. STAT 184 aims
to give you the theoretical and programming skills to understand and
develop RL algorithms. The course begins with simple environments
where you can solve for the optimal strategy exactly. Then it explores more
and more general cases that require more complex estimation and
optimization techniques. The homeworks are a mix of mathematical
problem solving and programming; There is a final project (no exam).

Statistics

Stat 110: Introduction to Probability

Stat 110 is what most people would consider to be the flagship course of
the Harvard statistics department, and for good reason— the material
you’ll learn in this class is fundamental not just for any future work you’ll
do in statistics, both theoretical and applied, but also in related
quantitative subjects like computer science, economics, engineering, and
mathematics. It is also by far the largest course in the Stats department,
enrolling over 500 students every fall.

For many students, this may be their first exposure to probability theory—
while statistics courses like AP Statistics and Stat 100 focus more on
practical applications of statistics for the purpose of data analysis, Stat 110
provides you with the theoretical probability toolbox that you’ll need to be
able to understand and quantify randomness and uncertainty in the world.

The majority of the content in Stat 110 revolves around understanding

various probability distributions that are commonly used in statistics and
data science, and the weekly problem sets and exams are designed to
challenge you and help you to develop the problem-solving skills and
mathematical maturity that will help you succeed in more advanced
statistics courses.
While prior experience with advanced math and probability can be
helpful, the only required background knowledge is single-variable
calculus at the Math 1b level and familiarity with matrices.

Stat 111: Introduction to Statistical Inference

After having taken Stat 110, you’ll hopefully be familiar with modeling
various random phenomena using random variables, which you can then
analyze using a variety of tools from the course. Stat 111 connects these
tools with data and the methods/ideas that most people associate with
“statistics”: hypothesis testing, confidence intervals, estimation, and
linear/logistic regression.

As opposed to Stat 110, most of the problem sets involve some coding in R
(you can also use Python, but there isn’t as much support). It’s not
necessary at all though to be familiar with R; if you’re familiar at all with
coding, you’ll pick up R skill through the class. GUSH also typically runs a
short R workshop before the class begins.

The main prerequisite for Stat 111, unsurprisingly, is Stat 110. The course
assumes deep familiarity with the content from Stat 110 (in particular,
details about the named distributions from Stat 110 and tools like Adam’s
law and Eve’s law).

Stat 117: Data Analysis in Modern Biostatistics

As the longest-standing and most "advanced" undergraduate biostatistics
course offered, Stat 117 benefits from both previous iterations of the
course and Giovanni's expertise in the field. Stat 117 runs as a 30-person
seminar style course where lecture notes and discussion prompts are
posted beforehand. Attendance is required but participation is informal
during the discussion-heavy classes—beware the 9am!

Content-wise, Stat 117 covers a wide range of biostatistical tools through

the lens of biomarker cancer research. All projects and biweekly
assignments center around two gene expression datasets; there are no
exams. Giovanni provides all of the limited biomedical background needed
and many additional resources and papers to understand the deeper
statistical underpinnings of tools used in the course.

Familiarity with Stat 110, 111, and 139 is helpful for understanding and
applying (in R) the tools in Stat 117.

Stat 139: Introduction to Linear Models

Stat 139 is the final course in the introductory statistics sequence of
110/111/139, and is probably the most applied out of the three. Like the
course title suggests, the majority of the course is devoted to understanding
the theory and applications of linear regression and related modeling
techniques (e.g. LASSO and Ridge regularization, variable selection, etc.)
and learning how to implement them in R. If you’re interested in doing
projects and research in applied statistics, Stat 139 will provide you with a
solid foundation in common modeling methods that you’ll use in fields
including but not limited to biostatistics, sociology, and economics.

Stat/CS 109a is another undergraduate-level course which covers a similar

set of topics. However, Stat 139 is more theoretical, has a larger emphasis
on linear regression and its variations, and assignments are completed in R
instead of Python.
Stat 210: Probability I
Stat 110’s older sibling. Stat 210 is the first graduate statistics course most
students take, and it’s usually also taught by Joe, so you can expect a fairly
similar lecture style. It’s run somewhat like an upper-level undergraduate
course: problem sets are still weekly, the class is large but is run like a
well-oiled machine, and there is a midterm and a final exam.

Notably, in contrast to analogous graduate probability courses at most

other universities, Stat 210 places greater emphasis on reasoning by
representation (i.e., learning the relationship between different common
distributions) instead of on measure theory, though measure theory still
comes up from time to time.

Make sure you’re solid on your Stat 110 content: familiarity with your Stat
110 distributions and stories, Adam’s/Eve’s law, LOTUS, etc. is assumed. .

Stat 211: Statistical Inference I

Stat 111’s older sibling. While Stat 111 introduces you to the principles of
statistical inference (likelihoods, estimators, hypothesis testing), Stat 211
dives into these topics on a much deeper and much more technical level.
You’ll prove a lot of the cool theorems taken for granted in Stat 111 (e.g.,
consistency and asymptotic normality of the MLE) and spend a lot more
time on hypothesis testing (e.g., how do you define and find the best or
most powerful test for a certain problem? what if we want to run lots of
tests at the same time?). Expect lots of Taylor expansions and tricky
homework/exam questions.
Stat 212: Probability II
Stat 212 is supposed to be a continuation of Stat 210, though it feels much
more like a standard math course than one of Joe’s statistics courses.
Content is covered quickly in lecture, but problem sets are typically
assigned biweekly. In Stat 212, you’ll work closely with martingales and
stochastic processes and potentially cover select advanced topics chosen by
the instructor.

Math at the level of real analysis is essentially necessary, and knowing some
measure theory from Math 114 can be super helpful.

Stat 213: Statistical Inference II

Stat 213 is a pretty natural continuation of Stat 211. You’ll learn when the
theorems in Stat 211 break down (e.g., when your likelihood function isn’t
“nice”) and powerful tools to use in these situations (e.g., quadratic mean
differentiability). You’ll also be introduced to a powerful new perspective
for asymptotic analysis: while any “decent” estimator/test should become
perfect as the sample size grows to infinity, if we make the problem
“harder” over time, certain estimators/tests seem to be “better.”

Stat 244: Linear and Generalized Linear Models

At a high level, Stat 244 combines the content of Stat 139 and Stat 149
(fittingly, 44 is also the average of 39 and 49). The first half of Stat 244
covers linear models, and the second half covers generalized linear models.
While these topics find common applications in industry, the focus of Stat
244 is primarily theoretical. The linear models aspect of Stat 244 also
differs from Stat 139 in this regard; while Stat 139 includes coding in R,
the biweekly problem sets in Stat 244 do not involve model
implementation.

The course listing recommends Stat 210/211 as prior background, though

concepts from these classes are not really used in the class. Stat 110, Stat
111, and familiarity with linear algebra is usually sufficient preparation for
the course.
The Concentration
As data grows in size, complexity, and importance, interest in statistics and
related yields has continued to grow. In 2008, Harvard’s Department of
Statistics had just 20 total concentrators; since 2018, we’ve consistently
had over 200 total concentrators.

The concentration itself is built upon the foundational courses Stat 110
(Introduction to Probability) and Stat 111 (Introduction to Statistical
Inference), which all concentrators are required to take for a letter grade by
junior year (they’re also often official or unofficial prerequisites for
electives). Stat 139, which teaches concentrators to implement and
interpret many frequently used statistical techniques, is also required for
all concentrators.1

The exact requirements of the concentration (linked here) will differ

slightly depending on which concentration track you pick. Ultimately, these
tracks don’t matter much: you can list them on your resumé, but it won’t
appear on your diploma.

1
It’s also possible to replace Stat 110 and/or Stat 111 with their graduate course
equivalents, Stat 210/211, though almost all students will take the undergraduate
version(s) before the graduate version(s). Similarly, it’s possible to replace Stat 139
with its graduate equivalent, Stat 244.
Planning Your Statistics Courses
There’s no single “correct” or “best” way, timeline, or pace to take statistics
courses at Harvard. Some people will come to Harvard with tons of math
experience and already determined to do statistics – which is great if that’s
the case – but so many (more than you’d expect!) eventual statistics
concentrators come to Harvard not yet ready to dive right into statistics
courses, whether it be because of their math background or their academic
interests at the time. This is 200% fine, and it is 300% possible to finish the
statistics concentration, even if you find later in college that you want to
concentrate in statistics. For the sake of hopefully providing some clarity,
we’ve listed some considerations for timing your courses, but these are in
no way authoritative, necessary, required, or even suggestions:

A great deal of timing statistics courses centers directly or indirectly

around Stat 110. Most concentrators take Stat 110 in the fall of their
sophomore year. Though there’s no need to rush through the course, it
can be nice to knock out Stat 110 once you’re ready because it’ll open up a
wide range of courses that list it (officially or unofficially) as a prerequisite:
e.g., Stat 111/171/210, CS 124/181, Ec 1011a/1126.

Freshman Year:
Freshman year can be a great time (a) to explore various courses related to
(and unrelated to!) statistics and (b) to take math courses to satisfy the
math requirement and prepare for statistics courses. There’s no need to
take a statistics course in your freshman year if it doesn’t make sense for
you, but some future concentrators will take:
● Stat 100. Stat 100 is an application-driven introduction to
statistics. While the course is not a required course for the
 concentration, it does (a) count toward the total number of
statistics courses required for concentrators, (b) give good practice
with R that can be helpful for Stat 111, and (c) help introduce
what statistics is. It is, however, very different from Stat 110 and
Stat 111 in terms of style and content – it should not be thought
of as a prerequisite for those courses.
● Stat 110. Some students with strong math backgrounds will take
Stat 110 their freshman year. If you’re one of these students, it’ll
give you a nice head start and give you early exposure to statistics,
but we want to reiterate that most students do not take Stat 110
until sophomore year or later.
● Math courses. Stat 110 will require some familiarity with calculus,
so it can be helpful for students to take Math 1a and 1b in their
freshman year if they are rusty or have not taken calculus in high
school. A proactive student can take Stat 110 with just Math 1a
(or Ma/Mb) under their belt: note that the calculus portion of
110 requires knowledge of integration and some knowledge of
series and Taylor expansions.

Students placed into Math 21+ will often take one of Math 21ab,
22ab, 25ab, 55ab: these satisfy the concentration’s two course
math requirements of linear algebra and multivariate calculus/real
analysis. Additionally, while Stat 110 doesn’t require formal
knowledge of proofs, having experience with proof-based
reasoning can be helpful for Stat 110.

It is also possible to satisfy the math requirements with other

courses. Some concentrators, for instance, will take Math 112
instead of 22b for a formal introduction to real analysis.
Sophomore Year:
Most concentrators – though, again, we want to emphasize this isn’t true
for all concentrators! – will take Stat 110 and Stat 111 in the fall and spring
of their sophomore year, respectively.

Junior and Senior Year:

The statistics concentration requires that both Stat 110 and 111 are taken
by the end of your junior year. Apart from those courses, students’
journeys through the statistics concentration varies a ton. A few
considerations:
● At this point, students still have the linear models (Stat 139)
requirement: there isn’t as much uniformity as to when students
satisfy this requirement.
● If you are considering pursuing the concurrent master’s program,
you will need to have satisfied the three course requirements for
the program by the end of the fall semester of your junior year.
Because of this, many of these students will take Stat 110 and 111
in their sophomore year and Stat 210 in the fall of their junior
year, though this is certainly not the only way that students satisfy
the concurrent master’s application requirements.
Sample Course Plans
Student starting in Ma

Freshman Sophomore Junior Senior

Fall Math Ma Math 1b Math 21b Stat 139

CS 50 Stat 110 Stat 108

Spring Math Mb Math 21a Stat 111 Stat 149

Stat 100 CS 51 Stat 171

Student starting in 1a

Freshman Sophomore Junior Senior

Fall Math 1a Math 21a Stat 139 CS 109a

Stat 110 Stat 99r

Spring Math 1b Math 21b CS 32 Stat 141

Stat 100 Stat 111 Stat 171 Stat 99r

Student starting in 21a, on the data science track

Freshman Sophomore Junior Senior

Fall Math 21a Stat 110 Stat 139

CS 50 CS 109a CS 61

Spring Math 21b Stat 111 Stat 171 CS 51

CS 181 CS 124
Student starting in 22a, completing the concurrent master’s program

Freshman Sophomore Junior Senior

Fall Math 22a Stat 110 Stat 210 Stat 211

CS 50 CS 109a Stat 139 Stat 244

Spring Math 22b Stat 111 Stat 220 Stat 288

CS 181

Student starting in 25a, accelerating through statistics courses

Freshman Sophomore Junior Senior

Fall Math 25a Stat 210 Stat 211 Stat 99r

CS 50 Stat 244
Stat 110

Spring Math 25b Stat 212 Stat 213 Stat 99r

Stat 111 CS 181
Concurrent Master’s (A.B./A.M.)
A handful of students each year (e.g., sixteen in the Class of 2024) are
enrolled in the concurrent master’s program, which allows students to
earn their bachelor’s degree (A.B.) and their master’s degree (A.M.) in four
years; that is, they’ll graduate with their undergraduate class but with an
additional master’s degree in statistics.

To enter the concurrent master’s program, students are required to

formally apply to the program at the end of their junior fall (rules are
slightly different for off-cycle students) through GSAS. In order to be
accepted, students must have completed at least three statistics courses,
which must include: (1) Stat 110 or 210 (with an A- or higher), (2) Stat
111 or 211 (with an A- or higher), and (3) at least one 200-level course
(with a B or higher) if not Stat 210 or 211.

For the application, you’ll need to fill out a plan of study, meet with Dr.
Glickman (the director of the concurrent master’s program in statistics)
and get two letters of recommendation. The two letters of
recommendation sound much scarier than they actually are: usually, the
statistics department just wants them to vouch for your ability to finish the
concurrent master’s program.

Statistics is not the only field in which you can pursue the concurrent
master’s program. A number of SEAS fields (e.g., computer science,
applied math, applied physics) have A.B./S.M. concurrent master’s
programs, and there are also a handful of other fields in GSAS (e.g., pure
math) that offer the A.B./A.M. concurrent master’s.
Research
Getting involved in research can be a super valuable and rewarding way to
augment the rest of your statistics journey at Harvard. Broadly speaking,
research is split into applied research and theoretical research (though this
split isn’t as clean as we might make it seem). Below, we outline what
applied and theoretical research look like, but we want to emphasize that
research – by its very nature – is highly variable, so your experience may
not perfectly match our descriptions.

Applied Research
Applied research often consists of writing code to clean, analyze, and
visualize complex datasets. Often, you’ll work with a PI (often a faculty
member), postdocs, and graduate students to accomplish data-related
tasks, such as producing tables and figures, implementing estimators and
algorithms, and performing exploratory analyses.

The process of applied research is often very iterative: you’ll write code to
accomplish some task, get feedback on what changes to your result might
be desirable, and then figure out how to adjust your code to accomplish
these changes.

Usually, for applied research, you’ll join a project as a member of a team:

the amount of control you have on the overall direction of a project can
vary by lab. Sometimes, you might be in charge of a subproject; sometimes,
you might jump around from different parts of the project, providing
coding support as necessary. That is, applied research can take the form of
being a research assistant (RA) for a professor, but it can also consist of
more independent work that is supervised by the professor.
Theoretical Research
Theoretical research is more geared toward proving certain statements
rigorously. This might involve proving lemmas as intermediate steps
toward proving a desired theorem/result. Often, you’ll work closely with a
faculty member who can guide you through intermediate lemmas and
provide advice for proving certain results.

Theoretical research can seem daunting at first, but in some ways, it’s not
too different from a problem set for a theory class. Usually, you’ll have
small things that you try to prove between each meeting with an advisor;
research meetings can then be used to think through tricky parts of a proof
with your advisor and review the general direction of a research project.

As opposed to a problem set, however, theoretical research projects usually

have much more lofty goals, so keeping the big picture in mind is
important while working on smaller intermediate proofs. Additionally,
while the claims that you’re asked to prove on a problem set are
(hopefully) known to be true, it’s not always immediately clear that a
certain lemma or theorem in a research project is true: because of this,
simulation, intuition, and heuristics play a large role in research. This part
of research – trying to figure out what things are actually true and what
aren’t – can be immensely frustrating but can also lead to new research
ideas and insights.

Theoretical research often has a higher barrier to entry than applied

research. Many advisors will often require that you have taken certain
courses to work with them, since their research might involve concepts
only covered in upper-level or graduate courses.
Finding An Advisor
Like the rest of this section, there’s no one way to find an advisor/lab.
We’ve included some considerations:
● The first step in finding an advisor/lab is often to figure out what
you’re interested in. For applied research, this might look like
finding an application area that you’re interested in (e.g.,
genomics, behavioral economics, or voting behavior). For
theoretical research, this might look like finding a subfield that
you’re interested in (e.g., free probability, high-dimensional
inference, conformal inference, etc.). You can definitely also shape
your interests around advisors/labs you would want to work with,
but it’s worth considering whether you’re really interested in the
relevant topics – it’ll make research a lot more engaging!
● There’s a few ways to find an advisor/lab. You might want to
work with a professor you’ve taken a class with, you might browse
through faculty members/labs on a department website, and
some departments (e.g., the Department of Economics or the
School of Public Health) have programs for matching students to
a project during different parts of the year.
● If there’s no formal process for matching students with a
project/advisor/lab, you’ll often have to reach out on your own!
You’ll often want to know some things about the
project/advisor/lab, both (a) to make sure you’re actually
interested and (b) to demonstrate that you know what you’re
signing up for. Then, you can reach out, asking if there are any
spots for undergraduate students (sometimes there aren’t!).
Funding and Getting Course Credit
Research can be great on its own, but it’s even better when you can get
paid or get course credit for your work! There’s a few different
opportunities for different parts of the year:

Faculty/Department Funding
A few faculty/labs are able to pay research assistants directly for the work
that they do, though this is more common in departments outside of
statistics and happens more often for applied research (e.g., many
economics professors pay their RAs on an hourly basis). Often, even these
groups will still ask undergraduate RAs to apply for funding (i.e., the
opportunities listed below).

Stat 91r
Stat 91r is one way to receive concentration credit for term-time research.
It must be taken as SAT/UNS and can count as one of the related courses
for the concentration. To “take” Stat 91r, you’ll need to already have a
faculty advisor, fill out the Stat 91r proposal form, receive permission from
Dr. Rader, and enroll in the course. There isn’t actually any formal course
content; rather, at the end of the semester, your faculty advisor will give
you a SAT or UNS grade.

Notably, you may not enroll in Stat 91r and be paid for your research work.
i.e., you cannot be enrolled in Stat 91r and receive a stipend from HCRP.

HCRP/Herchel Smith
HCRP is a research stipend offered by Harvard for term-time (up to
$1000) or summer research (up to $6000) supervised by a Harvard faculty
member. The HCRP application consists of a research plan (4-5
double-spaced pages) that outlines your project’s objectives, importance,
and methods and a letter of recommendation written by the faculty
member who will supervise your research.

Herchel Smith is a research stipend offered by Harvard for summer

research. As opposed to HCRP, Herschel Smith allows for a non-Harvard
affiliated faculty advisor and provides a slightly higher stipend. The
application is similar to that of HCRP but also includes a short essay
about how receiving the stipend will help you achieve your academic and
professional goals.

Research Village (PRISE/SPUDS/BLISS/KRANIUM/etc.)

The Harvard Summer Undergraduate Research Village (HSURV for
short) is a collection of different residential research programs hosted by
Harvard that run during the summer. Each program is focused on a
different discipline – SPUDS is focused on data science, BLISS is focused
on social sciences, etc. – and run slightly differently, but most programs
consist of research work with an advisor/mentor/lab, occasional
professional development/faculty presentations, and a final presentation at
the conclusion of the program.

Many of the programs will ask that students arrange advisors/labs/projects

themselves before submitting their applications (the specific project is
usually not binding), though a few programs have students apply to
specific projects.

Students receive room and board for the summer and a moderate stipend
(around $3000). HSURV puts a great emphasis on the community aspect
of the program: Harvard pays for various activities throughout the
summer that are separate from research (e.g., trips to amusement parks,
attractions, festivals), there are (required) opening and closing dinners, and
students all live in the same house for the duration of the summer (e.g.,
most years, all students live in Winthrop).
Acknowledgements
This guide itself is massively inspired by the SPS Guide to Physics and the
Harvard Chemistry Club’s Guide to Chemistry, and we can only hope to
provide a portion of the guidance that those manuals have given students
over the years.

This guide wouldn’t have been possible without the invaluable help from a
great number of individuals who volunteered their time to give helpful
feedback, write blurbs for classes, and provide suggestions.

GUSH takes responsibility for any mistakes made in this guide and
caution that much of the text in this guide is opinion and is based on our
own experiences during our times at Harvard; i.e., you might have a
completely different experience than us!

Go forth and study statistics!