If some variable has a pretty long tail, what can happen is that most of the values in that tail get removed as outliers when the data is split across other columns.
A solution is to detect this by comparing the variable between its 1dim and Ndim synthesis, and forcing a left/right stitch to preserve the 1dim distribution.
If some variable has a pretty long tail, what can happen is that most of the values in that tail get removed as outliers when the data is split across other columns.
A solution is to detect this by comparing the variable between its 1dim and Ndim synthesis, and forcing a left/right stitch to preserve the 1dim distribution.