Mean residual life function limits

812 Views Asked by At

In my paper about mean residual life functions and heavy tails I've come across the examples listed in the picture below.

I've managed to successfully calculate the limits when $M \rightarrow\infty$ for the Exponential and Pareto distribution myself using the equation $$\frac{\int_M^\infty {(1-F(z))dz}}{1-F(M)}$$ however I fail to understand how to calculate the remaining three (marked in yellow). I am new to probability so please bear with me. Thank you for your aid.

1

There are 1 best solutions below

1
On BEST ANSWER

First, let us write the formulas themselves. As I come from an actuarial background, my notation is slightly different: $$\overset\circ e_m = \operatorname{E}[T-m \mid T > m] = \frac{\int_{t=m}^\infty S(t) \, dt}{S(m)} = \int_{t=0}^\infty \vphantom._t p_m \, dt$$ where $T$ is the lifetime random variable, $S(t) = 1 - F(t)$ is the survival function of $T$, and $\vphantom._t p_m = S(m+t)/S(m)$ is the probability of survival of a life aged $m$ to age $m+t$. The key observation is that as $m \to \infty$, both the numerator and denominator tend to $0$, so L'Hopital's rule applies. Then by the fundamental theorem, and noting $S(\infty) = 0$: $$\lim_{m \to \infty} \overset\circ e_m = \lim_{m \to \infty} \frac{-S(m)}{-f(m)} = \lim_{m \to \infty} \frac{-f(m)}{f'(m)},$$ where $f$ is the density of $T$. Now all that remains is to choose a density: for a gamma distribution with shape $\gamma$ and rate $\alpha$, we have $$f(m) \propto m^{\gamma-1} e^{-\alpha m},$$ consequently $$f'(m) \propto (\gamma - 1)m^{\gamma-2} e^{-\alpha m} - m^{\gamma - 1} \alpha e^{-\alpha m} = m^{\gamma - 2} e^{-\alpha m} (\gamma - 1 - \alpha m).$$ Hence $$\lim_{m \to \infty} \frac{-f(m)}{f'(m)} = \frac{m}{\alpha m + 1 - \gamma},$$ and as $m \to \infty$, this is clearly $1/\alpha$ as claimed. A similar approach works for the other two distributions. If you remain stuck in the calculation, I will amend this answer to include those computations.