Change the indices of the vanilla and boxed correlationsum

[apbraun]

Changed the indices of the outer sum of the correlationsum to run from i = 1 to N instead of i = w + 1 to N-w. The normalisation has been adapted accordingly.

In all uses of the function for testing or plotting I haven't noticed any considerable slowdown.

Whether this is an improvement in precision is unclear, the fractal dimension the two variants find, differs a bit for small number of points (N < 1000) which is expected when 20 out of 100 or 200 points are discarded:

Note: This change only changes the more general q-order correlation not the optimised version for q = 2 hence I chose q = 1.99

[codecov-commenter]

Codecov Report

Merging #26 (bd49d88) into main (9aca116) will not change coverage. The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main      #26   +/-   ##
=======================================
  Coverage   92.17%   92.17%           
=======================================
  Files          17       17           
  Lines         677      677           
=======================================
  Hits          624      624           
  Misses         53       53           

Files Changed	Coverage Δ
src/corrsum_based/correlationsum_boxassisted.jl	90.11% <100.00%> (ø)
src/corrsum_based/correlationsum_vanilla.jl	96.73% <100.00%> (ø)

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more

[Datseris]

thanks anton. can you please explain what are the error bars in the bottom plot panel?

[apbraun]

They're just the s05 and s95 value from the retrieved from the slopefit function.