Things break. Well, let me put it more softly: everything has a finite lifetime.
Good thing is that mathematical statistics can help making light about that amount of time.
A problem I was dealing with recently made me think about how wonderful the exponential distribution is to model things that break.
Given the lifetime of an object by a random variable where is an unknown cumulative distribution function and represents its density function, we can define the risk function - that gives the intensity of the rupture - as .
What a data scientist really wants to know is the probability that something breaks after a time
Basically, what this ratio says
The formula above can be approximated with , that is the conditional density that something will break right after time .
And now comes the trick.
Imagine is a random variable from an exponential distribution (its density function is given by , ), then
that is the parameter of the exponential distribution indeed. This parameter is usually provided by the vendor of the object that is going to break. It can be a number or even an equation with some other parameters. The problem is how to determine from parameter .
And there we go:
Integrating both members of the equation above, we get
Therefore et voila’.
By knowing parameter it is possible to calculate the cumulative distribution function and therefore the density function (by its derivative) of the random variable that indicates the lifetime of the object we want to estimate.
Merry Christmas (oo)
Before you go
If you enjoyed this post, you will love the newsletter of Data Science at Home. It’s my FREE digest of the best content in Artificial Intelligence, data science, predictive analytics and computer science. Subscribe!