1

In PCA and Factor Analysis, there is the term loadings, which refers to factor loadings (onto the original variable).

Does the term (original) variable loading (onto the latent factor) exist?

user_anon
  • 827
  • 10
  • 20

1 Answers1

2

This is just a situation where terminology is being used interchangeably. It is more customary to say that a variable "loads on a factor" or principal component (PC), rather than a factor loads on a variable (PC). Specifically, however, it's the same thing. Loadings are correlation coefficients, essentially between an original variable $X$ and a factor or principal component $F$.

It's better to get into the habit of saying that an "original $X$ variable loads on a factor (or PC) with loading 0.65" (and not ever saying a factor(PC) loads on a variable).

  • Thanks, but just to be clear I have another question: The interpretation of the loading is: 1. The higher the loading of a PC, the more influence it has in the formation of the variable. 2. The higher the loading of a variable, the more influence it has in the formation of the principal component score. 3. Both ? So you say the answer is 1? – user_anon Sep 15 '18 at 05:14
  • If it matters, the longer post of mine regarding the question is: https://stats.stackexchange.com/questions/366970/fundamental-difference-between-pca-and-fa – user_anon Sep 15 '18 at 05:27
  • Your answer seems to be in opposition to [this](https://stats.stackexchange.com/questions/218600/why-are-regression-coefficients-in-a-factor-analysis-model-called-loadings)... – user_anon Sep 15 '18 at 12:19
  • The answer to the comment is 2. The higher the loading of a variable, the more influence it has in the formation of the principal component score. For option 1, factors and PCs don't create variables, since all the data comes from the variables. –  Sep 15 '18 at 17:42