优化问题:
x <- c(1,2,3,4,5,8,4,5)
y <- c(-2,-5,-3,1,5,3,-17,1)
alpha0 <- 0.01
Q <- function(alpha, x, y){
n <- length(x)
mu <- sum(x)
xi <- sum(y)
L1 <- (alpha-1)*xi-alpha*mu
-n*log(gamma(alpha))+n*alpha*log(alpha)
return(-L1)
}
Qaf <- nlm(Q, alpha0, x, y, hessian=T)
Qaf <- nlminb(alpha0, Q, x=x, y=y, hessian=T)
Qaf <- optim(alpha0, Q, x=x, y=y, hessian=T, method="CG")
Qaf <- optim(alpha0, Q, x=x, y=y, hessian=T, method="BFGS")
Qaf <- optim(alpha0, Q, x=x, y=y, hessian=T, method="SANN")
Qaf <- optim(alpha0, Q, x=x, y=y, hessian=T)
每次优化都会给出不同的参数结果和不同数量的 NaN 警告()。我可以重新参数化参数空间来解决这个问题吗?如何重新参数化参数空间并使用 R 代码解决这个问题?