| Details: |
The advancement of science has brought us to the era of massive data sets. Many of these data sets contain a large number of features with much lesser sample sizes. Such high dimension, low sample size (HDLSS) data have nullified the use of traditional statistical methods and have called for the emergence of new methodologies. In this talk, I shall discuss the use of some conventional and unconventional pairwise distances to construct statistical methods which can be adequately used in HDLSS situations. I shall mainly focus on the two-sample testing problem and discuss connections with classification and clustering. |