I'm working with some survival probabilities, and i've gotten kind of confused.

For simplicity, the death rate ( mu(x) ), where x is age in years, are considered constant in intervals of 1 year. In my assignment, the probability of surviving 1 year at age x is given as:

p(x,1) = e^(-integral[0,1] mu(x,1) ds). Since mu is constant in intervals of a year, then just e^( -mu(x,1) ).

But obviously, since mu(x,1) is constant in this one year, then the probability of surviving should also be 1 - mu(x,1)? But 1-x <= e^(-x).

What is the flaw in my calculation here?


1 Answer 1


You say the death rate is $\mu(x)$ where $x$ is age in years and that $\mu(x)$ is constant intervals of years. If we modify this statement and say that $\mu(x)$ is the probability density of death, then you are correct that $\mu(x)$ is also the 1-year death probability for a subject with age $x$. It would not be true that a survival probability is $\exp(-x)$. Note: a probability density is akin to the height of a normal "bell curve", and the area under the curve is considered a probability.

Another type of density that characterizes survival is the exponential with rate parameter $\lambda$. Exponential RVs are unique because they have a constant rate unlike Weibull. For exponentially distributed failure times, and if $\lambda=1$ then the survival probability beyond time $x$ is $\exp(-x)$.

One scenario that might characterize your confusion is if a survival regression model is used to model the $\lambda$ as a function of one or more regressors. In this case, you might use $\mu(x)$ to represent the model for the rate or intensity of the survival process, and age is just such an $x$ in the model, the 1 year survival probability for a subject with age $x$ would be $\exp(-\mu(x))$.

  • $\begingroup$ Thank you for the answer. I get that the probability of death is the area under the density. But in the case of mu being constant over a 1 year period, then the intergral over that 1 year, must simply be mu itself. So the probability of survival is 1-mu. I am a bit confused about the answer, though i appreciate it! Are you telling me that it is because the assignment assumes the probability of survival follows a different distribution? Thanks in advance! $\endgroup$ Commented Jun 16, 2023 at 12:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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