Re: OT: Incremental statistics functions

[tmoran@acm.org]

>I'm trying to find algorithms to calculate standard-deviation and variance
>_without_ keeping a buffer of the samples. I'm not really shure if I'm right
  I don't think I understand what you want.  You need only keep the
current count, sum, and sum of squares, as long as N is small and loss of
accuracy when adding a small number to a large sum isn't a problem.  If
that's all you need, I can send you my q&d regression package with spec:

package Regress is
  type Accumulator_Type is private;

  procedure Reset(Acc : in out Accumulator_Type);

  procedure Datum(X, Y : in     Float;
                  Acc  : in out Accumulator_Type);

  function N(Acc : Accumulator_Type) return Natural;

  function Avg_X(Acc : Accumulator_Type) return Float;

  function Avg_Y(Acc : Accumulator_Type) return Float;

  function Sd_X(Acc : Accumulator_Type) return Float;

  function Sd_Y(Acc : Accumulator_Type) return Float;

  function R(Acc : Accumulator_Type) return Float;

  function A(Acc : Accumulator_Type) return Float;

  function B(Acc : Accumulator_Type) return Float;
private
  type Accumulator_Type is record
    N : Natural := 0;
    Sumx, Sumy, Sumxy, Sumx2, Sumy2 : Float := 0.0;
  end record;
end Regress;