Jump to content

Talk:Scree plot

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Eigenvalue Criteria Cutoff of 1 ("Kaiser rule")

[edit]

In other places such as the ones below, it is often noted that the dimensions with eigenvalue less than 1 is not relevant to research. There are alternatives like CVE and PVE worth noting. Also the "Broken Stick Rule" is worth cross-referencing.

https://bradleyboehmke.github.io/HOML/pca.html https://www.graphpad.com/guides/prism/latest/statistics/stat_pca_graphs_tab.htm https://blogs.sas.com/content/iml/2017/08/02/retain-principal-components.html — Preceding unsigned comment added by 183.179.53.41 (talk) 15:39, 19 May 2022 (UTC)[reply]

I would like to add that the criticism section doesn't really work. Axis scaling, as long as it is linear, neither affects the eigenvalues nor moves the value of maximum curvature.
It may be reasonable to (properly) criticize using the plot as the only decision tool for picking (the number of) factors, but that does not mean it is a worthless view of data. Elias (talk) 05:47, 24 April 2023 (UTC)[reply]