Share your ideas for the JMP Scripting Unsession at Discovery Summit by September 17th. We hope to see you there!
Choose Language Hide Translation Bar
Level I

Missing Eigenvectors

Hopefully I'm not doing something silly, but I could use some help with some 'missing' eigenvectors. I have 10 specimens that in JMP table constitute 10 rows and 94 columns. Making the PCA and viewing eigenvectors is super simple. The output displays 9 PC columns with successive numbered Columns. And I see the variably colored values (darkest = presumably highest degree of variance?), BUT since I had 94 columns starting out, I would have presumed I would have had 94 vectors. But that's not the case - I have 46. Additionally, the eigenvector Column numbers are not in serial order. It's not simply 1-46; it starts at Column 5, a few teens are missing, skips around in the 30's, etc. I want to report the eigenvectors, so what am I missing? (figuratively and literally) Thanks!


Re: Missing Eigenvectors

I think some screen captures would REALLY help to see what you are seeing. Without that, here are a few of my guesses.


First, I am surprised that you are getting 46 PCs. I would have expected 9 or 10 since that is how many rows of data you have. Either way, I am guessing that you are not getting to the 94 expected PCs because you have already explained all of the variance. Go to the red triangle and ask for the Eigenvalues. Looking at that report, look at the last row in the Cum Percent column. Is it 100? If so, then you have explained all of the variance. Just because your data has 94 dimensions, it does not mean you have 94 dimensional data. In your case, it sounds like you have 46-dimensional data.


As for the original columns being listed in a different order on the eigenvalues, I would GUESS that you accidentally sorted the eigenvectors by one of your columns. Right-click on your eigenvector table and choose Sort by Column... and select the first column that contains your variable names. It will then sort in alphabetical order by that column.

Dan Obermiller
Article Labels

    There are no labels assigned to this post.