Subscribe Bookmark RSS Feed

PCA and variables significantly correlated to the PCs

r_shraim

New Contributor

Joined:

Apr 24, 2017

In PCA, is there any way to view or determine which variables are significantly correlated with the principal component axes? For example in the example that JMP gives ( https://www.jmp.com/support/help/13-1/Principal_Components_Report.shtml#213961 ) is there a table or something that gives the value of Ether's correlation with the PC1 or PC2 axis (r-value)??

Note: I'm using JMPGenomics - Thanks!

1 ACCEPTED SOLUTION

Accepted Solutions
mnelson

New Contributor

Joined:

Apr 18, 2017

Solution

Looks like Laura Higgins gives to information on how to infer what variables make up the PCs  watch her PCA webcast at https://www.jmp.com/en_us/events/ondemand/mastering-jmp/multivariariate-data-exploration.html I hope it help some.

3 REPLIES
mnelson

New Contributor

Joined:

Apr 18, 2017

I'm actually trying to figure out that too (if i understand your question properly).  Kemal Oflus has a PCA webcast at the following link.

https://www.jmp.com/en_us/events/ondemand/mastering-jmp/multivariate-analysis.html  In his webcast he shows that you can right click on the PC (once it's in your data table) and select "forumula" and it will show you the % of each variable that makes up the principal component.  In JMP 13 that does not work...  It seem there should be a way.  Judging by responses to most of these kinds of threads, Oflus must have used be black magic to get those percentages (or he was in error)...

Oflus WebcastOflus Webcast

mnelson

New Contributor

Joined:

Apr 18, 2017

Solution

Looks like Laura Higgins gives to information on how to infer what variables make up the PCs  watch her PCA webcast at https://www.jmp.com/en_us/events/ondemand/mastering-jmp/multivariariate-data-exploration.html I hope it help some.

r_shraim

New Contributor

Joined:

Apr 24, 2017

thanks! this is probably the best jmp can so far!