This is my code. The kum.loglik
function returns negative loglikelihood and takes two arguments a and b. I need to find a and b that minimize this function using optim function. (n1,n2,n3 is pre-specified and passed to optim function.
kum.loglik = function(a, b, n1, n2, n3) {
loglik = n1*log(b*beta(1+2/a,b)) + n2 * log(b*beta(1+2/a,b)-2*b*beta(1+1/a,b)+1) +
n3 * log(b*beta(1+1/a,b)-b*beta(1+2/a,b))
return(-loglik)
}
optim(par=c(1,1), kum.loglik, method="L-BFGS-B",
n1=n1, n2=n2, n3=n3,
control=list(ndeps=c(5e-4,5e-4)))
This code should work well but it gives error message
Error in b * beta(1 + 2/a, b) : 'b' is missing
What is wrong in this code?
The problem is (straight from the optim help):
fn: A function to be minimized (or maximized), with first
argument the vector of parameters over which minimization is
to take place.
Your kum.loglik
function needs to take a vector v
which you pull the parameters out of, e.g.:
kum.loglik=function(v) { a = v[1]; b = v[2]; ...}
I always use the following, it gives you the best results
p0 <- c(a,b) #example vector of starting values
m <- optim(p0, loglik, method="BFGS", control=list(fnscale=-1, trace=10),
hessian=TRUE, x=data.frame)
#for table of results
rbind(m$par, sqrt(diag(solve(-m$hessian))))
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With