什么情况下用LR 检验？

likelihood ratio test 中文应该是 似然比检验
suppose you have two model one restricted model and the other is a unrestricted model
under assumption: H0:theta=theta0 and H1: theta=theta1;
then by constructing two likelihood function L1(theta0|x)(unrestricted) and L2(theta1|x)(restricted) compute the likelihood ratio as LR(x)=L1(theta0|x)/L2(theta1|x)
reject H0 if {x<=c};
normally, we use -2ln(LR(x))~chi-squared(dim(Theta1)-dim(Theta0))