I'm developing a metamodel for my thesis and a sampling plan is required. Since i'm working with computer experiments I found Latin hypercube as one of the best options to design the sampling plan.
The point is that I work with 10 design variables and 66 points. However, because of the nature of Latin hypercube and its randomly work, I defined the design several times and obviously the obtained design was different every time (considering the same parameters and so). My question is if this is important, I mean, having different possibilities, and which one should I choose if in all cases the same criteria is imposed.
I hope you could help me to solve this thumb-problem, so any advice or
idea is welcome.