mantao/Plots.jl
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Improve histogram2d bin estimation

Browse Source 
The existing one would error when the correlation was high
This commit is contained in:
Michael K. Borregaard 2018-04-06 09:08:53 +02:00
parent 9acb89ba8a
commit 6ca8471c64
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/recipes.jl
View File

@ -562,7 +562,7 @@ function _auto_binning_nbins(vs::NTuple{N,AbstractVector}, dim::Integer; mode::S
# The nd estimator is the key to most automatic binning methods, and is modified for twodimensional histograms to include correlation
nd = n_samples^(1/(2+N))
nd = N == 2 ? nd / (1-cor(first(vs), last(vs))^2)^(3//8) : nd # the >2-dimensional case does not have a nice solution to correlations
nd = N == 2 ? min(n_samples^(1/(2+N)), nd / (1-cor(first(vs), last(vs))^2)^(3//8)) : nd # the >2-dimensional case does not have a nice solution to correlations
v = vs[dim]