Using Data Envelopment Analysis on survey data

Hi

I was looking to use DEA on some survey data but my sample size is quite low compared to my number of items. I have about 9 constructs with 50 or so items, and a sample of 22. I heard cross efficiency will help but how? I also could use some suggestions on software to use.

If anyone believes that DEA is not the way to go could someone provide an alternate form of testing relationships between two or more latent variables.

Thanks