Shiqin Helen Guan (grad student in statistics)

will give an introductory talk into robust statistics.
Very little prior knowledge is necessary for this talk,
and everybody is wellcome to join.

   Title:"Foundation for robust methods"

Abstract:

Conventional statistical methods for testing hypothese about mean and
regression parameters are based on the normality and homogeneity of
variance assumption. There is, however, a serious practical problem: these
standard methods can be highly unsatisfactory under the violations of the
model assumptions due to outliers and heavy-tailed distributions. Modern
robust statistical methods provide an effective way of dealing with these
problems.  Searching for robust measures which do not change dramatically
with slight changes in a distribution become the goal of many
statisticians. In this topic, we will discuss the foundations of roust
methods and the main tools of judging robustness.