Principal components interpretation

#1
I have principal components based on correlation matrix. Statistica PCA method report outputs not only factor loadings that depend on components' variances (eigenvalues)
but also contributions ( for example,
PC1=Lambda_11*X1+Lambda_12*X_2+...
so the contribution of jth variable to ith PC is squared Lambda_ij and visa versa X1=Lambda_11*PC1+Lambda_21*PC2+...)

For example, I have V1 and PC1 correlation 0.32 and V1 and PC2 correlation 0.33. And the first contribution is 1.5%
and the second is 4.1%.

Which of them should I use to interpret my PC ? Loadings that depend on
components' variances or just contributions ?
 
Last edited:
#2
Factor analysis

Forget correlations at this stage. Check the contribution of each variable to each component (factor). All those variables which are contributing more than 1( Kaiser's criterion) to the component are the part of the extracted dimension (factor). Look for the maximum contribution which is more than 1 and such variable is one you have to care about and relate to the new dimension.
 
#3
Pca

Forget correlations at this stage. Check the contribution of each variable to each component (factor). All those variables which are contributing more than 1( Kaiser's criterion) to the component are the part of the extracted dimension (factor). Look for the maximum contribution which is more than 1 and such variable is one you have to care about and relate to the new dimension.
Thanks for your reply. Does it mean that more than 1% contribution is
enough for variable to be related to PC dimension and be used for PC interpretation ? Is that correct ?
 
#5
Pca

So where do the factor loadings come out then ? If I use contributions for interpretation what are the correlations about
in this case ?
 
Last edited: