Open pierretoussing opened 4 years ago
@pawelbielski I am pretty sure there is no bug. I recreated the plots in 13_dependencies_to_different_levels
, but this time with the normal pearson's correlation and there you can see that in the equatorial region we have negative correlation.
As Pearon's and inverted Manhattan distance are nearly perfectly correlated, this plots definitely make sense.
The climate scientists are convinced there is a bug with the inverted manhattan distance when computing the similarities to different pressure levels.
Find the bug
If there is no bug, create plots to illustrate why we are seing this shifts in sign.