When: May 8–12, 2017. The course will run from 10-12 and from 12:30-14:30 every day.
Who: Jonathan Lilly
Application deadline: April 8. Registration form is found HERE. Members from CHESS/ResClim have priority. Registration deadline for participants from GEOMAR is February 20th.
Course credits: 2
The course is open to students from GEOMAR also. There are therefore limited space. The purpose of this course is to introduce students to the most useful tools and techniques for dealing with time series, with a focus on the particular needs of ocean/atmosphere research. Actual use of these methods in practice using Matlab, as well as their theoretical foundations, will both be emphasized. This year’s course will build on the version that was taught last year, the lecture notes for which are available at http://www.jmlilly.net/talks/oslo/index.html. Students who attended last year’s course, as well as those who did not, are both welcome. Whereas the first class emphasized a firm understanding of foundations, this course will focus more on a greater breadth of techniques as well as on practical applications. Beginning with the essential tools of skillful visual inspection, time-domain and multidimensional statistics, and bandpass filtering, we will progress to more advanced topics: Fourier spectral estimation, wavelet and time-varying spectral analysis and their underlying theories, covariance ellipse and polarization analysis, stochastic process modeling, and mapping of scattered data. For each topic, various example datasets will be analyzed and interpreted. In-class and homework assignments consisting of algebra, Matlab problems, and data analysis tasks will help the students solidify their understanding.
Outcomes: Students will gain a familiarity with a number of analysis methods, including new techniques that have been developed by the lecturer and colleagues specifically for ocean/atmosphere applications. Particular attention will be paid to clarifying common misunderstandings and avoiding potential pitfalls. In addition, students will learn hands-on by working in Matlab, primarily using the statistical and time series analysis toolbox jLab, written by the lecturer and available online at http://www.jmlilly.net/talks/oslo/index.html. As time series methods are, for historical reasons, often not particularly emphasized within the oceanographic community, it is hoped that this course will help students gain a fluency with tools that could prove highly useful to them in their future research.
Structure: The course will be held over the week of May 8–12, with a two hour session in the morning and another after lunch, with lectures and lab sessions roughly alternating. Problems and assignments will be given in order to help the students better understand the material. The final project will be a report illustrating the application of various techniques to a dataset of the student’s choice.
Requirements:Students are expected to have some prior experience with elementary statistics, complex exponentials, the Fourier transform, and the convolution theorem—the material covered in Chapters 3–6 in the lecture notes at http://www.jmlilly.net/talks/oslo/index.html. Each student is also expected to have a laptop running Matlab v2015a or later. Students are welcome to bring a time series dataset that they would like to analyze; otherwise, they will be able to choose from a number of sample datasets that will be provided.