2 ms·
The accepted draft of the research paper is in https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/2020GL091327 https://agupubs.onlinelibrary.wiley.com/doi
by gus_massa 6y ago
The accepted draft of the research paper is in https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/2020GL091327 https://agupubs.onlinelibrary.wiley.com/doi/epdf/10.1029/202...
Look at figure 10, graph O, P, Q in the last but one row. They are fitting the same data near 34 amu. The 3 vertical bars that show the error intervals are huge.
(The SAD values are in page 8, item 3.)
* In the graph O they fit the data using the expected curve of H2S+ and get a SAD value of 4.7
* In the graph P they fit the data using the expected curve of H3P+ and get a SAD value of 4.7 (again)
* In the graph Q they fit the data using the expected curve of of mix of H2S+ and H3P+ and get a SAD value of 4.0.
It's expected that with a mix you get a lower SAD, but it's clear that the error bars are so huge and the 3 fits so similar, that it's not possible to distinguish which one is the correct one. It looks like another case of overfitting.