NFYK15002U Advanced Methods in Applied Statistics

Volume 2019/2020
Education

MSc Programme in Physics
MSc Programme in Environmental Science

Content

The course will offer the practical knowledge and hands-on experience in computational analysis of data in all frontier physics research, with particular relevance for particle physics, astrophysics, and cosmology. The course content is based on statistical methods and does not require a specific or broad physics background. It is therefore applicable for many non-physics disciplines in the physical sciences.

Interested Ph.D. students and non-physics M.Sc. students in the Physical Sciences are very welcome to enroll.

Learning Outcome

Knowledge:

  • Be familiar with a supervised machine learning algorithm and multivariate analysis techniques, e.g. Boosted Decision Trees
  • Parameter estimation and uncertainty estimation using likelihood and Bayesian techniques
  • Minimization techniques using Markov Chain Monte Carlo and numerical methods

 

Skills:

  • Maximum Likelihood fitting
  • Construction of Confidence Intervals and contours
  • Code a chi-squared function in the language of the students preference (Python, C/C++, Ruby, JAVA, R, etc)
  • Creation and usage of spline functions
  • Application of Kernel Density Estimators
  • Inputing and processing data from both ASCII-readable files as well as internet data scraping.

 

Competences:

This course will help students develop the computational tools, software development, and use of statistical software packages for data analysis. The data analysis techniques are reinforced through assignments, which are import for  manuscript preparation, thesis writing, and understanding the methodology and statistical relevance of results in journal articles. The students will have enhanced general coding skills useful in the both academia and industry. Students will develop their own software solutions and tools and strengthen their independent problem solving skills.

Literature

No required literature.

 

For those looking for additional material, “Statistical Data Analysis” by G. Cowan is an excellent choice.

 

Class lecture notes and links to scholarly articles will be posted online.

- It is absolutely necessary to have extended knowledge and skill with at least one applicable computer programming language (Python, C/C++, Ruby, R, JAVA, or MatLab) for the course, with a preference for Python or C++. At a minimum, students should have accumulated at least 100 hours of writing, modifying, and debugging code in a single software language. If you have any questions or concerns about the coding competency required, please contact Jason Koskinen.
- The ability and experience to install external software packages, e.g. the MultiNest Bayesian inference package or “emcee” Markov Chain Monte Carlo sampler.
- Completion of “Applied Statistics: From Data to Results”, or equivalent, is strongly encouraged but not strictly required.

Academic qualifications equivalent to a BSc degree is recommended.
Instructor lectures, in-class examples, computer-based exercises, and discussion.
It is expected that students bring their own laptops or have access to a computer upon which they can install software to write, compile, and execute code.
  • Category
  • Hours
  • Lectures
  • 36
  • Practical exercises
  • 32
  • Preparation
  • 102
  • Project work
  • 36
  • Total
  • 206
Continuous feedback during the course of the semester
Credit
7,5 ECTS
Type of assessment
Continuous assessment
Written assignment, 28 hours
Assessment will be based on:
- An in-class short oral presentation (10%)
- Graded problem set(s) and project(s) centering around the coding, implementation, and execution of a statistical method (50%)
- Take home final exam (40%)
Aid
All aids allowed
Marking scale
7-point grading scale
Censorship form
No external censorship
Several internal examiners
Re-exam

Take-home exam in coordination with the course responsible, 28 hours.

Criteria for exam assesment

see learning outcome