By taking X to be a two-dimensional standardized random variable with zero mean and identity covariance matrix,
Question:
By taking X to be a two-dimensional standardized random variable with zero mean and identity covariance matrix, interpret Q4(W), defined in the previous exercise, as a directional standardized kurtosis. Taking ϵ1 as defined in Exercises 2.36 and 2.37, show, using the polar representation, that that ϵ2 − ϵ1 and ϵ4 − ϵ1 are both invariant under rotation of X, but change sign under reflection. Find expressions for these semi-invariants in terms of the κs. Interpret τ0 as the mean directional kurtosis and express this as a function of 4. Discuss the statistical implications of the following conditions:
16τ
2 1 = 9(κ30 + κ12)
2 + 9(κ03 + κ21)
2 16τ
2 3 = (κ30 − 3κ12)
2 + (κ03 − 3κ21)
2
(i) τ0 = 0;
(ii) τ2 = 0;
(iii) τ4 = 0;
(iv) ϵ4 − ϵ2 = 0.
Step by Step Answer:
Tensor Methods In Statistics Monographs On Statistics And Applied Probability
ISBN: 9781315898018
1st Edition
Authors: Peter McCullagh