Features TBD:

+ histogram(+ min, max, [rl]trim, n) returning [[ height, low, high ], ... ]
+ sample scaling (aka "forget gradually")
* clone( + min, max, zero_thresh, base, ... )
* cdf( $x, $y ) for intervals
* generate (random var with $self's distribution)
* remove_data( )
* total_abs_error, mean_abs_error - think more
+ linear interval around zero

Cleanup:

* more detailed docs on methods
* rationale

Features for discussion:

* better mode algorithm
* linear interpolation on add_data
* how to find the bulk of a distribution
* C++/XS library
* analysis tools (like: Your sample looks like 45% normal (100, 15) + 35% exp (100) + 20% utter rubbish)

Multidimentional S::D::LogScale:
* linear/log storage
* correlation for 2D
* all other stuff
