2
$\begingroup$

Suppose I have two random variables, $X_1$ and $X_2$, that are independent (but not identically distributed) and assume both have hazard functions $\lambda_1(s)$ and $\lambda_2(s)$, for $s > 0$. Then how can I derive the hazard function of $\min(X_1, X_2)$?

EDIT: Let $S$ describe the lifetime of a person, then define $\theta_s(t) = P(S > s+t | S > s)$ as the survival function. Then define the hazard function as $\lambda(s) = \lim_{t \rightarrow 0} \frac{1}{t} (1-\theta_s(t)) = -\theta_s'(0)$

$\endgroup$
  • 1
    $\begingroup$ Please define the term hazard function in the context that you are interested in. $\endgroup$ – Ric Apr 22 '14 at 7:26
  • $\begingroup$ Please see edit $\endgroup$ – SwiftMo Apr 23 '14 at 15:40
4
$\begingroup$

Note that $$P(X_i >s)= \exp\Big(-\int_0^s \lambda_i(u) du \Big),$$ for $i=1, 2$. Then, $$P(\min(X_1, X_2) >s) = P((X_1>s)\cap (X_2>s)) = P(X_1>s)P(X_2>s) = \exp\Big(-\int_0^s (\lambda_1(u)+\lambda_2(u)) du \Big).$$ That is, the hazard function for $\min(X_1, X_2)$ is $\lambda_1(s)+\lambda_2(s)$.

Alternatively, note that $$\lambda_i(s) = \lim_{\Delta s \rightarrow 0} \frac{P(X_i \leq s+\Delta s | X_i > s)}{\Delta s} \\ = \lim_{\Delta s \rightarrow 0} \frac{P(X_i > s) - P(X_i > s+\Delta s)}{\Delta s P( X_i > s)},$$ for $i=1, 2$. Then the hazard function of $\min(X_1, X_2)$ is given by $$ \lambda_{\min(X_1, X_2)}(s) = \lim_{\Delta s \rightarrow 0} \frac{P(\min(X_1, X_2) \leq s+\Delta s | \min(X_1, X_2) > s)}{\Delta s} \\ = \lim_{\Delta s \rightarrow 0} \frac{P(\min(X_1, X_2) > s) - P(\min(X_1, X_2) > s+\Delta s)}{\Delta s P( \min(X_1, X_2) > s)} \\ = \lim_{\Delta s \rightarrow 0} \frac{P(X_1 > s)P(X_2> s) - P(X_1 > s+\Delta s)P(X_2 > s+\Delta s)}{\Delta s P(X_1 > s)P(X_2> s)} \\ \small= \lim_{\Delta s \rightarrow 0} \frac{P(X_1 > s)P(X_2> s){\tiny-P(X_1 > s)P(X_2 > s+\Delta s) + P(X_1 > s)P(X_2 > s+\Delta s) }- P(X_1 > s+\Delta s)P(X_2 > s+\Delta s)}{\Delta s P(X_1 > s)P(X_2> s)} \\ \small=\lim_{\Delta s \rightarrow 0}\bigg(\frac{P(X_2> s)-P(X_2 > s+\Delta s)}{\Delta s P(X_2> s)} +\frac{P(X_2 > s+\Delta s)}{P(X_2> s)} \frac{P(X_1 > s)-P(X_1 > s+\Delta s)}{\Delta s P(X_1 > s)}\bigg) \\ = \lambda_1(s)+\lambda_2(s). $$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.