ecolmas-course2003-01

April, 15 - 17, 2003, Bremen

Introduction to statistical data analysis in earth sciences: uni-, bi- and multivariate methods

Tilo von Dobeneck, David Heslop (EUROPROX, RCOM)



Objectives
Every data set collected in earth science is an incomplete expression of one of the innumerable present or past experiments conducted by nature itself. It was shaped by an entangled system of physical, chemical and biological controls and laws including a generous pinch of chaos and random, not the least due to the decisions and measures taken by the investigator him- or herself. To discover these hidden truths and promises in limited and blurry information is our daily aim and duty in geoscientific research. But even intense gazing at our precious data sets doesn't always get us there...

Descriptive and analytical statistics offer a large variety of methods to investigate geological data in a systematic, quantitative and scientifically recognized approach. The theory and computation behind statistical methods may seem highly demanding (and often dull) to the non-expert, but their strategy, applicability and meaning in a given earth scientific context can usually be described in a practical and intuitive way. Thanks to existing powerful and user-friendly software packages, launching the most complex statistical techniques is now as simple as a mouse click - a chance as well as a danger for every eager practitioner.

This course will take a 'hands-on' approach to statistics by introducing and applying statistics bottom up. All exercises are exclusively based on fictive or real geological cases. We will learn how to mine data sets with one, two or more variables employing suitable uni-, bi- and multivariate methods. Please note that methods for time series analysis and data mapping will not be covered in this course.

Topics
- objectives, concepts and terminology of basic statistics
- analytics and interpretation of frequency distributions
- statistical hypothesis testing by parametric and non-parametric methods
- covariance and correlation, linear and non-linear regression
- cluster analysis and multidimensional scaling
- discriminant analysis, principal component and factor analysis

Textbook
I personally recommend 'Introduction to Geological Data Analysis' by Swan and Sandilands, Blackwell Science, 446 p., ISBN 0-632-03224-3

Sofware
Windows version of STATISTICA
PLEASE BRING YOUR OWN COMPUTERS / LAPTOPS, IF POSSIBLE!

Course schedule
Morning (9.00-12.30) and afternoon (13.30-17.00) sessions.

Location
Research Center Ocean Margins, University of Bremen, Germany
Am Fallturm 1, 28359 Bremen, Room TAB Seminar 1 (close to Geo Dep.)

To subscribe
mail to: Torsten Bickert ()