[Quasimarion]

Honestly the charts in the paper is not very convincing. The outliers are all over the place and there is hardly an association judging from naked eye.

(e.g: imagine seeing figure 4 without the model prediction in the background)

[wholinator2]

Agreed, it looks like they may have even missed a U-shaped quadratic association. If were doing pure curve fitting it'd work way better with those data. For real though, very unconvincing graphs.