P.C. MAHALANOBIS
CONTRIBUTIONS
Mahalanobis Distance
Mahalanobis Distance
is one of the most widely used metric to find how much a point diverges from a
distribution, based on measurements in multiple dimensions. It is widely used
in the field of cluster analysis and classification. It was first proposed by
Mahalanobis in 1930 in context of his study on racial likeness. From a
chance meeting with Nelson Annandale,
then the director of the Zoological Survey of India, at the 1920
Nagpur session of the Indian Science Congress led to Annandale asking him to
analyse anthropometric measurements of
Anglo-Indians in Calcutta. Mahalanobis had been influenced by the
anthropometric studies published in the journal Biometrika and
he chose to ask the questions on what factors influence the formation of
European and Indian marriages. He wanted to examine if the Indian side came
from any specific castes. He used the data collected by Annandale and the
caste-specific measurements made by Herbert Risley to come up with the
conclusion that the sample represented a mix of Europeans mainly with people
from Bengal and Punjab but not with those from the Northwest Frontier Provinces
or from Chhota Nagpur. He also concluded that the intermixture more frequently
involved the higher castes than the lower ones. This analysis was
described by his first scientific paper in 1922. During the course of
these studies he found a way of comparing and grouping populations using a
multivariate distance measure. This measure, denoted "D2"
and now eponymously named Mahalanobis distance, is independent of
measurement scale. Mahalanobis also took an interest in physical
anthropology and in the accurate measurement of skull measurements for which he
developed an instrument that he called the "profiloscope".
No comments:
Post a Comment