Re:Londoner solves 20k-year Ice Age drawings mystery
by jpivarski on Sunday January 08, 2023 @02:47PM
(#34302398)
It would be exciting if this is true. I just can't get past the statistical analysis.
Has anyone else noticed that the fit functions in Figures 3, 4, and 5 of https://www.cambridge.org/core/journals/cambridge-archaeolog... go right through the _centers_ of most of the histogram bins? It's pretty clearly a binned fit and the fit function is just picking a few bins to overfit to while ignoring the other bins. With so little data, it would be more robust to do a simple mean or median, rather than this misleading fit. Later, it looks like they're inferring p-values from the fitted parameters, which aren't going to mean very much.
Maybe the ultimate conclusions of this study are correct anyway. It just bothers me that the data analysis isn't justifiable as-is.
Has anyone else noticed that the fit functions in Figures 3, 4, and 5 of https://www.cambridge.org/core/journals/cambridge-archaeolog... go right through the _centers_ of most of the histogram bins? It's pretty clearly a binned fit and the fit function is just picking a few bins to overfit to while ignoring the other bins. With so little data, it would be more robust to do a simple mean or median, rather than this misleading fit. Later, it looks like they're inferring p-values from the fitted parameters, which aren't going to mean very much.
Maybe the ultimate conclusions of this study are correct anyway. It just bothers me that the data analysis isn't justifiable as-is.