CS 1380 + ORIE 1380 + STSCI 1380
Data Science For All
Spring 2018
Catalog Description: This course provides an introduction to data science. Given data from economics, medicine, biology, or physics, collected from internet denizens, survey respondents, or wireless sensors, how can one understand the phenomenon generating the data, make predictions, and improve decisions? We focus on building skills in inferential thinking and computational thinking, guided by the practical questions we seek to answer. The course teaches critical concepts and skills in computer programming and statistical inference, in conjunction with handson analysis of realworld datasets, including economic data, document collections, geographical data, and social networks. We will also consider social issues in data analysis such as privacy and design.
Lecture: MWF 10:1011:00 am in Kimball B11, starting January 24, 2018.
Instructors: Michael Clarkson and Madeleine Udell
Topics and Course Objectives
A schedule of lectures and assignments is available.
This course is positioned at the intersection of computing and statistics, with an emphasis on empirical analysis of realworld data sets through computation, rather than mathematical theory. There are no prerequisites other than highschool algebra. As the Cornell motto says, we welcome “Any person…any study.”
The course is organized into three units:

Exploration: discovering and visualizing patterns in data. Basic Python programming, arrays, tables and table manipulation, bar charts, scatter plots, line graphs, and histograms.

Inference: drawing conclusions from data, and quantifying the reliability of those conclusions. Experiments, randomized controlled trials, association, causation, laws of probability, probability and empirical distributions, sampling, estimation of parameters, variability, mean, median, standard deviation, the normal distribution, the Central Limit Theorem, hypothesis testing, pvalues, confidence intervals, the bootstrap.

Prediction: making guesses about unknown quantities based on data. Correlation, linear regression, least squares, minimization, residuals, nonlinear regression, multiple regression, dummy coding, nearestneighbor classification, training vs. test data, accuracy.
Outcomes:

Students will learn to communicate data visually and to interpret visualizations for themselves; and will practice interpreting their discoveries in nontechnical terms to people who are not data scientists.

Students will analyze more than two dozen realworld data sets, from entertainment, sports, demographics, medicine, banking, transportation, voting, real estate, and sustainability.

Students will become familiar with Jupyter notebooks, which are a standard tool for analyzing and presenting data; will learn simple procedural programming in Python; and will learn to use the
datascience
Python package for table manipulation and plotting. The package is based on and simplifies the usage of Pandas and Matplotlib, which are standard packages for data science, while providing the possibility for a smooth transition to those packages in students’ future work. 
Students will learn to conduct hypothesis tests by correctly formulating null and alternative hypotheses, choosing test statistics, and empirically computing pvalues through simulation.

Students will learn to estimate population parameters, such as the mean and median; and to use the bootstrap to compute confidence intervals.

Students will learn some basic properties of the normal distribution, including how to describe it with mean and standard deviation, standard units, a simple statement of the Central Limit Theorem, and how to choose sample sizes based on the square root law.

Students will learn to use linear regression to predict quantitative values, and to calculate and interpret the correlation coefficient, slope, and intercept of the regression line; and will learn to compute linear and nonlinear regression models through numerical minimization of leastsquares error.

Students will learn to apply nearestneighbor classifiers to predict categorical and quantitative values, and to implement such classifiers from scratch; and will learn some basic aspects of machine learning, including splitting data into training and test sets, and measuring the accuracy of a classifier.
Answers to Some Important Questions
Q: Is it ok if I’m undeclared? Or if I’m majoring in something other
CS, ORIE, or Stats?
A: Yes! All majors are welcome, especially those from outside
CS, ORIE, and Stats!
Q: Do I need to know a lot of math for 1380?
A: Basic high school mathematics (e.g., Algebra I and II) is all you need.
We won’t use any calculus in 1380.
Q: Do I need to know how to program for 1380?
A: Nope! We’ll teach you everything you need to know.
Q: Can I take 1380 if I’ve already taken a class on introductory programming
(e.g., CS 1110) or stats (e.g., AEM 2100, ENGRD 2700, HADM 2010, ILRST 2100,
MATH 1710, PAM 2100, STSCI 2100)?
A: Yes, but if you’ve taken both programming and stats before,
you’re likely to find 1380 to move too slowly for you.
You could instead consider INFO 2950, INFO 3300, CS 4780, ORIE 4740, or STSCI 4060.
Acknowledgment
This course is based on Data 8, a course taught by Ani Adhikari and John DeNero at the University of California, Berkeley. They and their teaching assistants have developed many of the materials we are using in our own course. We are using those materials with their permission, which we gratefully acknowledge.