Suppose we have a vector of numbers, and say we also have a null distribution for this vector (say, we pick a vector of numbers from a multivariate gaussian density); what would be the best way to estimate the deviation of the vector of numbers from the null multivariate gaussian distribution? (let's also say that we didn't know that the null was multvariate gaussian, that is, we only had an empirical multivariate distribution of vectors that we did not know a priori conform to any particular distribution, then what is the best way?)
 
thanks,
p

Reply via email to