Meetings: Tuesdays and Thursdays, 1-2pm, BH 317
Website: http://jfukuyama.github.io/teaching/stat710
| Instructor: Prof. Julia Fukuyama | jfukuyam at iu dot edu | |
| Office hours: Thursdays 2-4pm | Office: Informatics East, Room 201 | |
| Associate Instructor: Mr. John Koo | johnkoo at iu dot edu | |
| Office hours: Mondays 10:30-11:30am and Wednesdays 1-2pm | Office: Informatics East, Room 103 |
As a statistician, you will need to manipulate data, optimize, and simulate. You will also need to know enough about how the methods you use work to diagnose problems when they arise and to be able to implement modified versions when the standard implementations don’t suit your purposes.
You also need to write accurate, clean, maintainable, demonstrably correct code. To that end, the first half of the class will be devoted to how to program well, with statistical tasks giving us the computational problems. The class will be primarily in R, with one homework in python and the option of another homework in python for those who would like more experience with the language.
Once we have the software engineering down, we will move on to the algorithms used in applied statistics. These can be roughly broken up into optimization methods and stochastic simulation methods. For optimization, we will cover gradient descent, stochastic gradient descent, the EM algorithm, and topics in convex optimization. Stochastic algorithms will include rejection sampling, Metropolis-Hastings, and Gibbs sampling.
The primary textbook for the course with be The Art of R Programming, by Norman Matloff.
The R Cookbook, by Paul Teetor, will also be useful.
Additional readings will be posted on the course website.
Assessment will be based on a combination of homework, an in-class midterm, and an in-class final on the scheduled final exam date. Final grades will be based on:
There will be 10 homeworks over the course of the semester, generally graded out of 5 points, with one point for a good-faith effort at all the problems, 5 points for correct answers with clean code, and an intermediate number of points otherwise.
Homeworks will be assigned on Sundays and due the following Tuesday (9 days later). At the time the homework is assigned, we will generally not have covered all the material needed to complete the homework, but we will have covered everything by the Thursday before the due date. The idea is to give you the homework early enough that you can think about it while the material is being covered in lecture. Therefore, it will generally be a good idea to take a look at the homework when it is assigned even if you aren’t able to complete all the problems yet.