The name is absent

36

3.2 Why copulas and which copulas?

3.2.1 Advantages of the copula approach

Among many approaches in the sensitivity analysis, the copula approach is partic-
ularly appealing to our research. One of the advantages is that by using copula
functions, the joint distribution of two variables can be obtained without the require-
ment of knowledge of the marginal distributions. Such attribute is naturally suitable
in estimating the joint distribution of events and dependent censoring, a critical step
to derive the conditional probability of survival (See Section 3.3.2 for details).

Another advantage of copulas that benefits this research is their flexibilities. Ex-
tensive copula functions have been well studied (Nelson, 2006, and see Section 2.3
for details). The wide range of copula functions serves versatile needs on modeling
joint distributions of events and dependent censoring. In addition, some fundamental
research of applying copula functions to study financial data has been published re-
cently (Li, 2000; Cherubini et al, 2004), which provides further insights into the ways
of utilizing flexibilities of copulas.

3.2.2 Which copula function to use?

Previous studies demonstrate that, in contrast to choosing an incorrect copula func-
tion, falsely assuming the value of Kendall’s τ leads to a much more severe bias in
parameter estimates. For instance, setting τ^,s value as zero (assuming independent

<<   46   47   48   49   50   51   52   >>

More intriguing information