Alternative distribution to exponential distribution (negative power distribution)?

1.5k Views Asked by At

I'm recently trying to understand where the exponential distribution is originating from, and while learning the exponential distribution, I started exploring whether or not a similar alternative distribution would have some practicality in real world practice.

The exponential distribution pdf is $f(x) = k\mathrm{e}^{-kx}$, and its cdf on $[0, \infty)$ is equal to $1$ for any $k$.

The distribution I came up with has a pdf of $f(x) = k(x+1)^{-k-1}$, and its cdf on $[0, \infty)$ is $1$ for any $k$.

Both functions are decreasing to represent a decreasing pdf at greater values of $x$, but what are some reasons why people would prefer an exponential model over a power model?

1

There are 1 best solutions below

0
On

Exponential distribution has some very nice property, like memoryless property, relation with Poisson process inter-arrival time, etc, and the distribution itself is relatively simple. It can be used as a toy model to understand the problem.

https://en.wikipedia.org/wiki/Lomax_distribution Whereas the Pareto distribution (Power Law distribution) in general has a heavy-tailed, which used to model those extreme event, and is useful among those, say insurance problem. Many real-world data are actually heavy-tailed and using light-tailed distribution for modelling will create a modelling risk as it cannot reflect the tail behavior. You may also take a look at the term "Black Swan".