Forecasting has been used to predict elections, climate change, and the spread of COVID-19. Poor forecasts led to the 2008 financial crisis. In our daily lives, good forecasting ability can help us plan our work, be on time to events, and make informed career decisions. This practically-oriented class will provide students with tools to make good forecasts, including Fermi estimates, calibration training, base rates, scope sensitivity, and power laws. We'll discuss several historical instances of successful and unsuccessful forecasts, and practice making forecasts about our own lives, about current events, and about scientific progress.
Prerequisites: One of Stat 134, Data/Stat C140, EECS 126, Math 106, IND ENG 172, or
equivalent; and familiarity with Python (Pandas); or consent of instructor.
Strongly Recommended: Compsci 61A, Data/Compsci C88C, or equivalent.
Stat 165 and Stat 265 will share the same lectures, but the assignments and work will be different.
Weekly homework and forecasting exercises will be assigned in both classes but homework in Stat 265
will assume a greater degree of mathematical maturity.
For Stat 165 students, there will be a quiz at the beginning of class on most Fridays, which covers the reading from that week. It will generally be short (1-2 multiple-choice questions). The two quizzes where you got the lowest grade will be dropped. Students in Stat 265 will not be assigned quizzes. On Fridays, students in both Stat 165 and Stat 265 will have an in-class worksheet due that is graded for completion.
Instead of exams, both classes will require a final project. The final project for Stat 265 students will be more in-depth and research-focused. Students in Stat 265 will propose a research project to the staff for approval. The main criterion is that the project should consider some novel facet of forecasting such that it could in principle be extended to a publishable paper. For instance, some past projects involved a novel analysis of scoring rules for forecasts, and a user interface study for forecasting software.
All students are given up to 5 total slip days across the semester to turn in homework late. It is up to the student to notify course staff in advance if they plan on using slip days for homework 1 (unless they have DSP accommodations) since not completing homework 1 on-time will result in you getting dropped. These slip days are intended to be used for extenuating circumstances, so we recommend students only use them when necessary. After that, assignments are penalized 20% per day and drop to 0 after 3 days late. For Friday worksheets, if a student contacts us in advance with a reason for missing discussion, they can complete the worksheet on their own and turn it in online. For weekly forecasting exercises, since they concern current events that happen shortly after the assignment is due, we are unable to accept late submissions. However, students will be given extra credit opportunities to make up for missed assignments not covered by the policies above.
Please email firstname.lastname@example.org if you need any particular accommodations, even if you have already requested services on your DSP portal.
You can join the EdStem here: https://edstem.org/us/join/rJxBKt.
MW4-5PM, in Anthro/Art Practice Bldg 160
F4-5PM, in Wheeler 212
This class will be heavily disussion-based and participation will count towards the grade. Monday and Wednesday lectures will be a combination of traditional lecture and group activities, while most Fridays will be student-led small group discussions with instructors helping to facilitate.
Instead of exams, there will be a final project. Students in Stat 265 will be expected to do a more substantial project.
There will be no official lab / discussion block, but some homework will involve Pandas programming.
Our office hour schedule this semester will be:
- Jacob Steinhardt (Lead Instructor): 325 Evans, Wednesday 3-4pm
- Matthew Dworkin (TA): 428 Evans, Thursday 12:30-1:30pm; Friday 3-4pm
- Rahul Shah (TA): 428 Evans, Tuesday 11am-Noon; Wednesday 1-2pm
Grades will be based on a combination of:
Grades will be based on a combination of:
As a baseline, we expect that the forecasts that ChatGPT gives to result in roughly a B- on the forecasting competition section of your grade.
Steinhardt, J. (2023). Forecasting
To reach course staff, you can email email@example.com. If possible, please avoid emailing professors or GSIs directly!