Some stochastic comparison results for frailty and resilience models

Arindam Panja,Pradip Kundu,Biswabrata Pradhan

DOI: https://doi.org/10.48550/arXiv.2209.08261

2022-09-17

Abstract:Frailty and resilience models provide a way to introduce random effects in hazard and reversed hazard rate modeling by random variables, called frailty and resilience random variables, respectively, to account for unobserved or unexplained heterogeneity among experimental units. This paper investigates the effects of frailty and resilience random variables on the baseline random variables using some shifted stochastic orders based on some ageing properties of the baseline random variables. Relevant examples are provided to illustrate the results. Some results are illustrated with real-world data.

Statistics Theory

What problem does this paper attempt to address?

The problem that this paper attempts to solve is in survival analysis and reliability engineering, how to more accurately model the influence of heterogeneity on individual hazard rates or reverse hazard rates by introducing random effects. Specifically, the paper studies the influence of frailty and resilience random variables on the baseline random variable, and uses some shifted stochastic orders based on the aging properties of the baseline random variable for comparison.

### Main problems:
1. **Frailty model**: Study the influence of the frailty random variable on the baseline random variable, especially under different frailty distributions and baseline distributions, how to use the shifted stochastic order to compare the modified random variable \(X^*\) and the original baseline random variable \(X\).
2. **Resilience model**: Similarly, study the influence of the resilience random variable on the baseline random variable, especially under different resilience distributions and baseline distributions, how to use the shifted stochastic order to compare the modified random variable \(X^*\) and the original baseline random variable \(X\).

### Specific objectives:
- **Definition and properties**: Define the shifted stochastic orders (such as the upward - and downward - shifted likelihood ratio order, hazard rate order, reverse hazard rate order, remaining life order, etc.), and discuss the application and superiority of these orders in stochastic comparison.
- **Theoretical results**: Derive the influence of frailty and resilience random variables on the baseline random variable under different assumptions, including results in terms of likelihood ratio order, hazard rate order, reverse hazard rate order, and remaining life order.
- **Practical applications**: Verify the validity of the theoretical results through two practical data sets (leukemia survival time and bearing fatigue life data).

### Key formulas:
- **Survival function of the frailty model**:
\[
\bar{F}^*(t)=\int_{0}^{\infty}\bar{F}(\lambda t)dH(\lambda)
\]
- **Distribution function of the resilience model**:
\[
F^*(t)=\int_{0}^{\infty}F(\lambda t)dH(\lambda)
\]
- **Definition of shifted stochastic orders**:
- Upward - shifted likelihood ratio order \(X\leq_{\text{lr}^\uparrow}Y\):
\[
\frac{g(x)}{f(x + t)}\text{ is increasing with respect to }x\text{, for all }t>0
\]
- Upward - shifted hazard rate order \(X\leq_{\text{hr}^\uparrow}Y\):
\[
\frac{\bar{G}(x)}{\bar{F}(x + t)}\text{ is increasing with respect to }x\text{, for all }t>0
\]
- Downward - shifted likelihood ratio order \(X\leq_{\text{lr}^\downarrow}Y\):
\[
\frac{g(x + t)}{f(x)}\text{ is increasing with respect to }x\text{, for all }t>0
\]
- Downward - shifted hazard rate order \(X\leq_{\text{hr}^\downarrow}Y\):
\[
\frac{\bar{G}(x + t)}{\bar{F}(x)}\text{ is increasing with respect to }x\text{, for all }t>0
\]

Through these studies, the paper aims to provide a more refined method to understand and quantify the influence of heterogeneity on hazard rates in survival analysis and reliability engineering.