I have seen that some people are talking about "factor loadings" in PCA. It is a topic that I do not manage to understand, despite some googling.

I managed to obtain some code that generates the following plot (for 9 different "elements"). But I cannot interpret the results.

enter image description here


I believe your two plots are factor loadings given by PCA for the first two principal components.

  • The bar represents the magnitude for each variable "loaded" on the latent component
  • The bar also represent whether the loading is positive or negative

Based on the plots, I can see variable 4 and 6 are highly loaded on PC 1. Thus, we can say the variable 4 and 6 are similar on the PC 1. There must be something in the original data set that makes variable 4 and 6 similar. Note that all loadings on PC 1 are positive

We can also see variable 6 continues to dominate the PC 2. This is a variable you might find useful for modelling. A few variables are now negatively loaded on PC 2, but they are all less than 0.5.

Although variable 4 and 6 are highly loaded on the first PC, they are not exact copy of each other.

Variable 7, 8 and 9 might not be very important for your analysis because they have low factor loading on both plots.

When you do PCA, you should also get the variance explained by each PC.

  • 1
    $\begingroup$ When you say "factor loadings", do you mean "PCA eigenvectors"? $\endgroup$ – amoeba May 8 '17 at 15:06

Not the answer you're looking for? Browse other questions tagged or ask your own question.