Michael was confused how the correlation could be e.g. -0.38 when this point is not in the threshold graph at all. This is not very strange, however, because the threshold correlation only measures what happens in the first and third quadrants.
To make this more accessible to the reader, I think we should either explain very explicitly (note that the correlation overall NOT has to be in the range of the threshold) or include an adjacent scatter plot, as Christoffersen does. Scatter might crowd the paper but maybe it's good.
Michael was confused how the correlation could be e.g. -0.38 when this point is not in the threshold graph at all. This is not very strange, however, because the threshold correlation only measures what happens in the first and third quadrants.
To make this more accessible to the reader, I think we should either explain very explicitly (note that the correlation overall NOT has to be in the range of the threshold) or include an adjacent scatter plot, as Christoffersen does. Scatter might crowd the paper but maybe it's good.