Hi all,
I am using the function COV.WT to estimate the estimators (location and
scale) of a bivariate cauchy distribution.
My doubt is about the option WT (weight), cause at the R-help shows that the
weight is uniform according to the number of observations. But, checking the
theory, for example, the mean is given by
mean_estimator=mean(u(s)x)/mean(u(s)), where
x=my data (bivariate)
u(s)=(df+p/df+s), df=degree freedom, p=number parameters, s=mahalanobis
distance
The u(s) function is considered as a weight, in which gives more weight for
the observations with a small mahalanobis distance.
My question is, should I use this function u(s) as my weigth option at the
COV.WT function.
Is that clear!
Thanks in advance,
Marc