Yes, I think you can choose the component with highest variance. The first component is already a combination of different original vectors.
Normally, we choose the first couple of components which can explain more than 90% of variance.
you can try R data miner. It is easy to see variance contribution.
Does this help?