Here's my data. Note the "outlier"

The correlations are

X vs Y: 0.39

ln(X) vs Y: 0.32

But if I remove the outlier, the correlations DROP

X vs Y: 0.15

ln(X) vs Y: 0.08

I am used to seeing the correlation increase when removing outliers. In fact, that's often (usually??) a reason for removing outliers, no?

Now, I am not a fan of removing outliers, and I am not sold on removing this one. And I really don't want to get into that discussion here (whether or not to remove outliers). But I would like to know when it is common to see correlations decrease after removing outliers like this one. Anyone have a good mathematical explanation?

Skip