I'm aware of a general rule of thumb that says to take at least 10 times as many samples as there are parameters, but are there any issues to consider for Latin Hypercube sampling as the number of parameters increases?
Specifically, I would like to sample the parameter space of ~30 input parameters for a model using a Latin Hypercube (currently a conservative estimate!). I'm not particularly limited by computing time for this model so I'm more interested as to whether there are issues with efficiently sampling the parameters space.
Any comments or references to papers would be greatly appreciated!
Yes. The rule of thumb is that you need to have at least ten times more observation then you have number of parameters in the model.
Thanks for your response Besnik. Do you know if that rule still holds for 30-40 parameters?
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Sampling Design