[statistics] Updated abstract test for distributions

[Alex Herbert <al...@gmail.com>]

I have created a revised design for the abstract test for the distributions.

The current design requires that a single parameterization of the
distribution is created and this is then tested for the PDF, CDF,
SurvivalFunction, and InverseCDF. There are also tests for consistency
of survival and CDF, sampling, density integrals, and support bounds.

If you wish to test another distribution you can create one and call
various test functions but not all of them. You cannot repeat the
tests for consistency of survival and CDF, sampling, density
integrals, and support bounds.

So you cannot fully test additional distributions.

I have created a new design where all tests are performed on
distributions that are created dynamically by streaming arguments for
their construction and then streaming the test input and expected
values. So far I have applied this to 1 distribution which is in the
following PR:

https://github.com/apache/commons-statistics/pull/30

Look at the full file for the Nakagami distribution and it should
clarify that testing individual parameterizations of distributions
does not require much code.

I propose to change over the distributions to the new design and add
additional test cases for many of the distributions which have a
different PDF shape with different parameters.

Alex

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org