Principal Components Analysis (PCA)



Hello everyone. :blush:
I am a new user of Rsudio, and I need some help.
I doing a work with data from six countries (for 3 months), and seven variables for study as: manufacturing Energy, Capital goods…
My teacher compels us to use the “Statis” and “prcomp” functions.

Then we have to reserch new “Packages” for graphics and PCA analysis to cross data
For the moment I already found a difficulty. I use “biplot” and “epPCA” for comparing and with these two I find solutions differents. it is as if the graph is inverted relative to the point center of the axes

I do not know what I’m doing wrong or if this situation can happen and have solution!
I would like some tips from the community for solve this problem, and ask for new packages/functions that i can use to improve my study,
Thank you in advance for your attention.
Greetings for all :slightly_smiling_face:


Welcome Pika_78!

Firstly, I’m curious if you could repackage your question into code that I and other’s here can reproduce on our own machines? This will help enormously in addressing your question about finding different solutions.
To do this, check out reprex.

I don’t personally have much experience with PCA myself, but I think r-bloggers has a few articles on the topic.


I don’t know much about the STATIS technique, but isn’t it a slightly different take on PCA? E.g.

Does your prcomp biplot look more like the second set?


The sign in principal component analysis is arbitrary, since PCA is a simple mathematical transform of your data. When you flip the direction along the axis, the variance doesn’t change, and neither does the interpretation.

You can find some great answers at CrossValidated that explain this, e.g: