They say that fitting different degree Polynomials robustly identifies inflexion points seeing 10, 30 and 80. I think they only went for a higher degree because it was “visually underfit” at lower degrees, ie no scientific basis.
This is already after the dimensionality reduction which has its own arbitrary choice that affects what inflexion points you can identify.
This definitely smells like “we threw some data into a bunch of statistical analysis without thinking about it and wrote down anything that looked publishable”
They say that fitting different degree Polynomials robustly identifies inflexion points seeing 10, 30 and 80. I think they only went for a higher degree because it was “visually underfit” at lower degrees, ie no scientific basis.
This is already after the dimensionality reduction which has its own arbitrary choice that affects what inflexion points you can identify.
This definitely smells like “we threw some data into a bunch of statistical analysis without thinking about it and wrote down anything that looked publishable”