I'm trying to create an environmental composite index in which i have 22 variable under 7 dimensions. Due to limited sample size of 25, a component wise PCA was implemented. But since variables are not much correlated, the KMO values are coming very less, what can be done?
If the variables are not so much correlated a single composite index has no 'natural' meaning, thus you need to use more than one index. If your 7 dimensions carry each one a 'recognizable' meaning you could use 7 indexes and if for some reason you need 'a summary' you could compute a global averaga across the 7 indexes (after a standardization). But this is 'imposed' by your needs and does not come up by the correlation structure of the variables.
Dear Poorvi,
Thank you for your question, I advice you to read my new article as shown below:
https://doi.org/10.1016/j.jtemb.2021.126733
I have discuss the PCA in details.
Best,
Baker
I developed a while ago an index that I called RQE correlation index
Please find the link here https://www.ias.ac.in/article/fulltext/conf/001/01/0035-0041
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling