Bathtub curve

[Dario Bianchi]

Using the library GSL of C++ I tried to construct the bathtub curve like the sum of a gamma distribution (first part) and a weibull distribution (second part). I succeed in resolving the integral of this sum, but when I plot the hazard rate I found a bathtub, but after the rising part on the right the curve start decreasing and oscillates until become zero. Is it a problem of the algorith that GSL uses or something else?

[Tom Anderson]

If you used an increasing failure rate model (for Weibull this would be a beta > 1), then at some point all the population would have been exhausted to failure and you would find, prior to this time, a maximum failure fraction and after this a decrease in the fraction -it may fail faster, but there is less and less to fail from the population.

[RS Support]

No, once all the population has failed the failure rate (hazard rate) may start approaching zero. Another way to look at it is by saying what is the mortality rate for 200 year old humans...