0

I have a set of 100 variables and aim to reduce my data dimension for further subsequent analyses. There are about 300 observations.

Upon prcomp() in R (with retx= TRUE), my PC1's loadings (abs) are all below 0.20. And, most other PCs' loadings (abs) are also in the range of 0.01-0.30.

How would you do in this situation?

Thank you.

Joe
  • 8,073
  • 1
  • 52
  • 58
mchsu
  • 31
  • 1
  • 5

1 Answers1

2

Please make sure you scale the data before PCA. If you don't then the loadings can get skewed. Set the scale variable in prcomp to 'TRUE'.

Aditya Lahiri
  • 409
  • 3
  • 11
  • Don't forget the dot: `scale. = TRUE` – Joe Dec 06 '18 at 13:07
  • Yes, I have. My code looks like this: prcomp((pca.data[,202:304]^(1/3)), center= T, scale=T,retx= T). None of my PCA loadings is greater than 0.20. – mchsu Dec 07 '18 at 14:43