Many statistical procedures and techniques are based on a set of assumptions, such as normal or other parametric distributions, independence, homogeneity of variance, etc. However, what if some of those assumptions are not true, or the assumed distribution is mis-specified? This question leads to a fascinating and active area in modern statistics, called nonparametric statistics, which was developed to overcome the limitations of parametric methodologies. Nonparametric methods aim to conduct inference for the underlying population with few or minimal assumptions made on the population distribution or model structure, yielding higher degree of flexibility and robustness. This course covers both classic nonparametric methods that are based on ranks, signs, or permutations, as well as modern parametric techniques, including nonparametric cumulative distribution estimation, nonparametric kernel density estimation, nonparametric regression, bootstrap and jackknife resampling schemes, selection of smoothing parameter (cross-validation), among others. Throughout the semester, students investigate these methodologies and implement them to simulated or real datasets using the statistical language R. Prior experience using R is expected.

Units: 1

Max Enrollment: 15

Prerequisites: MATH 205, STAT 260 or STAT 318, and MATH 220/STAT 220.

Distribution Requirements: MM - Mathematical Modeling and Problem Solving

Typical Periods Offered: Every other year

Semesters Offered this Academic Year: Fall

Notes: