Phys 515 Spring 2010 --- Data Analysis (Phys 515)

Phy 515

Back to SEOS

The main course website is on moodle.

Matlab

We recommend Matlab for the second half of the course. Matlab is a proprietary interpreted language like python or perl that has an excellent graphical interface. You can write functions, scripts, or explore data from the command line, and this

Matlab is installed on various UVic computers, including the Clerihue labs. If your computer runs an X11 server, you may also run it via ssh -X unix.uvic.ca. From the command prompt type matlab and if ssh is working properly you will get a matlab environment on your remote computer. You may need to enable X11 Forwarding - google this if you do not know how it should be done.

Windows users will need to download and install an X11 server.

Nik Zappantis can help with real problems, but please ask your classmates or labmates for help before bugging Nik as he is very busy.

Matlab has an excellent tutorial as part of it's documentation.

Assignment 3

Assignment 3 is now available...

You will need a data file, available here:

• ShearData.mat

This file is binary, so be sure not to convert to ascii. If you would like this as a non-Matlab format you can use the text files at:

• ShearData.txt.zip

Handing this in:

Please hand in your m-files and a short PDF document summarizing your results. More figures are good. You will not be critiqued on your writing style, and point-form is OK. However, completeness and clarity is important so we can tell that you have understood the assignments.

To hand in, please tar cfz or zip your files (please don't use any other archive formats) into one directory with your lastname and the class id: i.e. LastnameP515/, and name your archive the same way (i.e. LastnameP515.tar.gz). Name your pdf file the same way, (i.e. LastnameP515.pdf). No need to uniquely identify your m-files so long as they are in a directory.

Email your archive to mailto:jklymak@uvic.ca.

PLEASE DO NOT INCLUDE THE DATA FILES (we have them already!).

If you used tar you should have an output like:

> tar tfzv LastnameP515.tar.gz
>   LastnameP515/LastnameP515.pdf
>   LastnameP515/mfile1.m
>   LastnameP515/mfile2.m
>   ...

Spectral Analysis: (3 Lectures)

• Reading:
  • Some notes summarizing the math from class can be found here. Please let me know if there are any mistakes so I can correct for next year.
  • Bendat and Piersol, 1986: Random Data; Analysis and Measurement Procedures
  • Emery and Thompson, 2004: Data Analysis Methods in Physical Oceanography
• Tutorial: Make some data to do spectral analysis on
• Tutorial: Take a raw power spectrum of some data
• Tutorial: Some drawbacks of spectral analysis
• Tutorial: Techniques for overcoming drawbacks
• Tutorial: 2-D spectral analysis
• M-files: A tar-gzipped package of m-files that contain the source code for the tutorials used above.

Outline

• Discrete Fourier Transform
  • Fast-Fourier Transform in Matlab
• Example transforms
  • Delta-function
  • White noise
• Pitfalls:
  • Spikes
  • Aliasing
• Periodigrams:
  • Windowing
  • Statistical significance
    • degrees of freedom
    • spectral smoothing
  • Trade-offs.
• Co-Spectra:
  • Coherence
  • Phase
• 2-D spectra

Principal Component Analysis

Not lectured on since 2008. This material is here for your abundant spare time...

• Reading: Shlens, 2005; This intro by Slens is a very informative discussion of how PCA works and what it is useful for.
• Tutorial: Two-dimensional PCA
• Tutorial: Large-dimension PCA: North Pacific

Outline

• Other names: EOF, SVD
• Why use PCA?
  • determine dominant modes of variability
  • reduce degrees of freedom of datasets.
• Linear Algebra
• Examples:
  1. bivariate in time
  2. spatially variable

Last Modified: March 3, 2010

---

Jody Klymak - jklymak@uvic.ca