Displaying 2 results from an estimated 2 matches for "kbll2".
2012 Jul 17
0
Maximum Likelihood estimation of KB distribution
...()*, I first declared the
Negative Log-likelihood of this distribution as follows;
*Loglik.newdis2<-function(x,a,b,n,imax) {
term<-0
for (i in 0:imax) {
term=term+(((-1)**i)*(choose(b-1,i))*(beta(x+a+a*i,n-x+1)))
}
dens=a*b*choose(n,x)*term
KBLL2<-sum(log(dens))
return(-KBLL2)
} *
since there is an infinite convergent series in this PMF, I decided to
specify a maximum value as imax instead of infinity without loss of any
information, and n is the binomial trials.
Please tell me whether the declared negative loglikelihoo...
2012 Jul 11
0
declaring negative log likelihood of a distribution
...a large
number like 1000 is defined as the maximum
#with for loop
*>Loglik.newdis2<-function(x,a,b,fre,n,imax) {
term<-0
for (i in 0:imax) {
term=term+(((-1)**i)*(choose(b-1,i))*(beta(x+a+a*i,n-x+1)))
}
dens=a*b*choose(n,x)*term
KBLL2<-sum(fre*log(dens))
return(-KBLL2)
} *
#Now to estimate the parameters, I used
http://rgm2.lab.nig.ac.jp/RGM2/func.php?rd_id=bbmle:mle2 mle2(from the
package bbmle) , which is a wrapper around the optim function
*>estimates1=mle2(Loglik.newdis1, start=list(a=1,b=1),
da...