Santhosh Kumar Rajamani
Usually 3, less than that you are essentially doing Statistical test instead of PCA.
0 votes
0 thanks
David Morse
Hello Amina,
The key question here is, for what application?
1. For creating a new set of linearly independent composite scores? One variable.
2. For data reduction? Two.
3. For generating subsets of affine variables in order to best account for overall variance via some systematic structure? Two. (Note: if there's any possibility of measurement error and/or variance specific to individual variables, common factor analysis would be a preferable choice to PCA.)
4. For testing in measurement/SEM models? Three. (See note in #3 as well.)
5. For something else? You'll have to explain further.
Good luck with your work.
0 votes
0 thanks
- Badges
- Science topic
More Amina Ali's questions See All
Similar questions and discussions