MVP Exponential Distribution Test
The MVP Test for Exponentiality, which assumes an unknown origin parameter, employs the following statistic:
The expected value of this test statistic is approximately one, for sampling from an exponential distribution. For most non-exponential distributions, the value will be much greater than one. Note though, a Chi-Square distribution, with one degree of freedom, will have an expected value less than one.
The significance (p-value) of this test is generated through simulations as follows.
- One million simulations are taken from a theoretical exponential distribution.
- The significance is obtained by taking the proportion observed in the simulations above the test statistic.
- A two-tailed significance value is generated.
- After 20,000 simulations, if no values, or all values, fall above the test statistic, the p-value is estimated to be zero.
- The program provides an abort button, to stop the p-value generation.