Interpreting PCA with VARIMAX

[Traian]

I performed PCA with a varimax rotation on my dataset and obtained this.

Now i need to group the variables based on their correlations to these factors. My question is how to properly accomplish this?

My approach:

After looking into this i found that you want to keep the values >0.32 or <-0.32 that show strong correlation with only 1 factor. Therefore i'd have to scratch all variables that show up on multiple factors and work with the rest. I applied a fuzz=.32 to my PROC FACTOR function to easier see them.

Where does the 0.32 come from? I have read multiple articles on how to interpret these results and while most said values should be over 0.32 i have seen examples like 0.4 as well.

[PaigeMiller]

Here's a paper https://hosted.jalt.org/test/PDF/Brown31.pdf that references the original paper that put forth the ±0.32 limit, but I don't have access to the original paper.

[Traian]

Thank you for the article, i continued to read the next chapters to find out how to deal with the complex values(the ones with >0.32 for more than 1 factor). Unless i missed something, the author suggested proceeding with the higher value and ignoring the other one also higher than 0.32, or simply ignoring the variable.
In my case, all these variables are a type of product and i have to calculate the probability that a customer will buy the products that belong to each factor value. Looking at ia1_17, the values for Factor1 and Factor6 are way too close. I am having a hard time thinking i have to ignore this variable since i will only be left with 2 values remaining for Factor 6 which is way too small.