6
$\begingroup$

Let $X$ be a positive, continuous random variable. Denote the density of $X$ by $f(x)$ and its CDF by $F(x)$. Let $\bar{F}(x) = 1- F(x)$ be the survival function of $X$. Given that the Hazard Rate, \begin{align} \lambda(x) &= \frac{f(x)}{\bar{F}(x)} \end{align} is increasing, i.e. $\lambda'(x) \geq 0$, I want to prove that \begin{align} \mathbf{E}\left[ X-c \mid X>c \right] &\leq \mathbf{E}\left[ X \right] \end{align} for any constant $c >0$.

Here's what I have tried so far:

Let \begin{align} \Lambda(x) &= \int_0^x \lambda(s) \mathop{}\!\mathrm{d} s. \end{align} This implies the following \begin{align} \lambda(x) &= \Lambda'(x), \\ -\Lambda(x) &= \log \left(\bar{F}(x) \right), \\ \bar{F}(x) &= e^{-\Lambda(x) }. \end{align} Writing the expected values as integrals (using the Darth Vader rule) and using the above, I can rewrite the inequality as \begin{align} \int_c^\infty e^{-\int_c^s \lambda(x) \mathop{}\!\mathrm{d}x } \mathop{}\!\mathrm{d}s &\leq \int_0^\infty e^{ -\int_0^s \lambda(x) \mathop{}\!\mathrm{d}x } \mathop{}\!\mathrm{d}s. \end{align} However, I do not see how this holds given $\lambda$ in increasing, and I can't think of a way using that assumption.

$\endgroup$
2
  • 2
    $\begingroup$ You can pull a factor of $e^{-\int_0^c\lambda(x)\,\mathrm{d}x}$ out of the integral on the left, so the integrand is the same on both sides. I don't know if this helps, though. $\endgroup$ Commented Jun 24, 2020 at 17:24
  • 1
    $\begingroup$ I have played around with this idea for a bit and I could not get anywhere. It seems like I always end up in the same situation: I have 2 integrals, of which either the smallest has a factor greater than 1 multiplied, or the greater integral has a factor smaller than 1 multiplied. In either case, I am stuck. $\endgroup$ Commented Jun 25, 2020 at 9:38

1 Answer 1

Reset to default
3
+50
$\begingroup$

Since $\lambda$ is increasing, we have $\lambda(y+c) \geq \lambda(y)$, hence $\int_0^t \lambda(y + c) \,dy \geq \int_0^t \lambda(y)\,dy$, and thus $e^{-\int_0^t \lambda(y + c) \,dy} \leq e^{-\int_0^t \lambda(y) \,dy}$. Changing variables and using this inequality gives $$\int_c^\infty e^{-\int_c^s \lambda(x) \,dx} \,ds = \int_0^\infty e^{-\int_c^{t+c} \lambda(x) \,dx} \,dt = \int_0^\infty e^{-\int_0^t \lambda(y+c) \,dy} \,dt \leq \int_0^\infty e^{-\int_0^t \lambda(y) \,dy} \,dt $$ as desired.

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.