Abstract:
This article is an attempt to show how robust regression, a computer
based statistical technique introduced by P.J.Huber in 1973 and later developed by
Rousseeuw (1984), Rousseeuw and Yohai (1984), and many others, can helps us in
cases where OLS totally fails due to outliers, leverage points and non-normality of
error distribution. To infer from the estimators obtained from robust regression we
generally need, especially for small samples, bootstrapping (resampling) technique
that is also a computer intensive statistical technique introduced by Efron (1979),
and later developed in many directions. This talk illustrates the whole thing by an
example using data extracted from the Big Mac. Index with a purchasing power
parity analysis.
