Instructor: Sergei Kuznetsov, Math 318, Phone: 2-4579 (don’t leave messages, send an e-mail instead), e-mail: Sergei.Kuznetsov@colorado.edu, Office Hours: MWF 2-2:50 p.m.
Lectures: MWF 3-3:50 p.m. in ECCR 151
Text: Lecture Notes: http://euclid.colorado.edu/~sek/4540-15.pdf (updated 1/26/15, some typos fixed)
Topics to be covered: Chapters 1-7 from the notes. That includes basic properties, trend-based models, seasonal models, stationary series, ARMA and ARIMA models, spectral analysis, estimation of the power spectrum and cross-spectrum, frequency filtration, and forecasting with time series models.
Prerequisites: Intro to Probability + Intro to Statistics (Math 4510/4520 or equivalent). [Things like Calculus 1-2-3 go without saying]. We also need: (a) some Linear Algebra, like matrices and matrix multiplication, inverse matrices, determinants, and (b) some Complex Numbers (check out Appendix C in the notes).
Homework: There will be weekly assignments. Homework will include theoretical problems as well as numerical examples (data sets for them will be posted on the web). Some data sets are several hundred points long, so some computer skills required. You may use any software like Excel, Matlab, Mathematica, Maple or similar where you design the algorithm (so you may not use specialized statistical software like R and so on). As a rule, numerical problems can be easily done in Excel. For numerical problems, please attach the code to the homework.
Warning: Homework assignments are not of the same size, some of them may be 10 problems long while the others are only 2 problems long. Expect about 60-65 theoretical and 15-20 numerical problems overall. All problems will be graded, solutions will be posted on the web. Don’t skip assignments, especially big ones! Skipping a big one could cost you up to 10 % of the overall score.
Late Homework Policy: No penalty if you have a documented excuse. Also, you may ask for an extension either in person or by e-mail before the due date. The extension will be granted only twice during the semester (the class is way too big). No homework will be accepted if it is more than one week late or if solutions are already posted.
Examinations: There will be a take home Final.
Grading policy: Homework 70%, Final 30%
5540 students: Students enrolled in Math/APPM 5540, will have to do extra “in-depth” homework problems (expect 20 to 25 extra problems). They will have extra “in-depth” problems on the final as well.
Just in case, more reading: Chris Chatfield, The Analysis of Time Series, An Introduction, Chapman & Hall, 5th edition (very good, however skips some important details); Peter Brockwell and Richard Davis, Introduction to Time Series and Forecasting, Springer, 2nd edition (graduate level text, focuses on math models and like); M.B.Priestley, Spectral Analysis and Time Series, Academic Press (out of print since '96 or so; very good, balanced and very detailed graduate level text, 900+ pages long).