Department of Statistics, University of Manitoba

“Integrated Rank-Weighted Depth”

Date: Thursday, February 14, 2019

Data depth is an important nonparametric concept of data centrality. Different measures of data depth exist, most of them being extremely hard to compute in the case of high dimensional data - the very case where they are most needed. In this talk, we will introduce the concept of data depth and focus on integrated rank-weighted (IRW) depth, a very natural extension of univariate ranks to the multivariate setup. We will discuss some of the properties of this depth measure (including asymptotics) which has the advantage of being easily approximated through a simple Monte Carlo experiment, even in the context of very high dimensional data. We will also discuss the associated IRW median, defined as the point that is most central to a data set, or "deepest" point.

Important Dates

February 18: Louis Riel Day (University Closed)

February 18 – February 22: Reading Week (No classes)

Upcoming Seminars

Statistics seminar: Aleeza Gerstein: “Levelling up R for Statistical Research and Teaching (Vignettes from RStudio::Conf)” — Thursday, February 28 at 2:45 p.m., P230 Duff Roblin.

Statistics seminar: Jame Fu: “Distribution of number of Levels in [s]-specified random permutation: A finite Markov Chain Imbedding Approach” — Thursday, March 7 at 2:45 p.m., P230 Duff Roblin.

Statistics seminar: Alison Gibbs: “Teaching Statistics in a Data Science World” — Thursday, March 21 at 2:45 p.m., P230 Duff Roblin.

Where are they now?

John Petkau, B.Sc. Mathematics & Statistics (1971)

Mostofa Sarkar, M.Sc. (2012)