Abstract:In this paper, we discuss the worst-case of distortion riskmetrics for general distributions when only partial information (mean and variance) is known. This result is applicable to general class of distortion risk measures and variability measures. Furthermore, we also consider worst-case of weighted entropy for general distributions when only partial information is available. Specifically, we provide some applications for entropies, weighted entropies and risk measures. The commonly used entropies include Gini functional, cumulative residual entropy, tail-Gini functional, cumulative Tsallis past entropy, extended Gini coefficient and so on. The risk measures contain some premium principles and shortfalls based on entropy. The shortfalls include the Gini shortfall, extended Gini shortfall, shortfall of cumulative residual entropy and shortfall of cumulative residual Tsallis entropy with order $\alpha$.
What problem does this paper attempt to address?
This paper attempts to solve the problem of how to estimate the worst - case distortion riskmetrics and weighted entropy when only some information (such as mean and variance) is known. Specifically, the paper explores how to quickly estimate riskmetrics when only partial information of the data set is known, especially for the worst - case estimation of general distributions.
### Background and Motivation of the Paper
1. **Importance of Distortion Riskmetrics**: Distortion riskmetrics play an important role in the construction of premium principles and risk aversion, and are widely used in behavioral economics and risk management.
2. **Limitations of Existing Research**: Most existing research mainly focuses on the worst or best cases of general - class distortion riskmetrics (such as VaR, TVaR, and RVaR), while less involved in variability measures (or entropy) in distortion riskmetrics such as Gini, EGini, CRE, CT, and h - CRE.
3. **Requirements of Practical Applications**: In practice, for a set of data, usually only partial information (such as moment information) is known, and it is necessary to quickly estimate riskmetrics. Therefore, how to estimate the worst - case of entropy, weighted entropy, and riskmetrics when only the mean and variance are known is a common problem.
### Main Contributions
1. **Worst - Case of Distortion Riskmetrics**: The paper derives the worst - case of distortion riskmetrics for general distributions under the condition that the mean and variance are known.
2. **Worst - Case of Weighted Entropy**: The paper further extends the above results and derives the worst - case of weighted entropy for general distributions under the condition that the mean and variance are known.
3. **Application Examples**: The paper provides specific application examples of entropy, weighted entropy, and riskmetrics (including some entropy - based premium principles and shortfalls).
### Technical Details
- **Convex and Concave Hulls of Distortion Functions**: To describe the distributions in extreme cases, the paper defines the convex hull \(g^*\) and concave hull \(g_*\) of the distortion function \(g\).
- **Main Theorems**: The paper derives the upper bounds of the worst - case of distortion riskmetrics and weighted entropy by using the modified Schwarz inequality and Cauchy - Schwarz inequality.
- **Specific Examples**: The paper discusses in detail multiple specific examples of entropy and riskmetrics, such as cumulative Tsallis past entropy, cumulative residual Tsallis entropy, extended Gini coefficient, etc., and gives the analytical solutions of these metrics in the worst - case.
### Conclusion
The paper successfully solves the problem of how to estimate the worst - case of distortion riskmetrics and weighted entropy when only some information (such as mean and variance) is known. This result is not only of great theoretical significance but also has wide applicability in practical applications, especially in the fields of financial risk management, insurance pricing, etc.