401-3632-00L  Computational Statistics

SemesterSpring Semester 2015
LecturersM. Mächler, P. L. Bühlmann
Periodicityyearly recurring course
Language of instructionEnglish


Abstract"Computational Statistics" deals with modern methods of data analysis (aka "data science") for prediction and inference. An overview of existing methodology is provided and also by the exercises, the student is taught to choose among possible models and about their algorithms and to validate them using graphical methods and simulation based approaches.
ObjectiveGetting to know modern methods of data analysis for prediction and inference.
Learn to choose among possible models and about their algorithms.
Validate them using graphical methods and simulation based approaches.
ContentCourse Synopsis:
multiple regression, nonparametric methods for regression and classification (kernel estimates, smoothing splines, regression and classification trees, additive models, projection pursuit, neural nets, ridging and the lasso, boosting). Problems of interpretation, reliable prediction and the curse of dimensionality are dealt with using resampling, bootstrap and cross validation.
Details are available via Link .

Exercises will be based on the open-source statistics software R (Link). Emphasis will be put on applied problems. Active participation in the exercises is strongly recommended.
More details are available via the webpage Link (-> "Computational Statistics").
Lecture noteslecture notes are available online; see
Link (-> "Computational Statistics").
Literature(see the link above, and the lecture notes)
Prerequisites / NoticeBasic "applied" mathematical calculus and linear algebra.
At least one semester of (basic) probability and statistics.