#1




HW5  Isaac
I'm confused about multivariate chisquare. Is our chi square model (the one we wish to minimize):
, given data vectors, ? Our data is already Gaussian, but we need to generate an empirical chisquare for multivariates... 
#2




So in lecture 10 there's a slide that says that the chisquared is the above formula when the entries of data vector x are normal but not necessarily independent. What we need is to get something in the form
chi^2 = sum (yi) where the yi's ARE independent. Here's the strategy: 1. Use the fact that Sigma is positive definite and use Cholesky decomposition to put the RHS into the form mentioned. 2. Prove that those yi's are independent Siavash did a good job with the proof. 
#3




Thanks Courtney.

Thread Tools  
Display Modes  

