R help - Sep 2008 - Error in "[<-"(`*tmp*`, i, value = numeric(0)) :

I use "while" loop but it produces an errro. I have no idea about
this.

Error in "[<-"(`*tmp*`, i, value = numeric(0)) : 
        nothing to replace with

The problem description is

The likelihood includes two parameters to be estimated: lambda
(=beta0+beta1*x) and alpha. The algorithm for the estimation is as
following: 
1) with alpha=0, estimate lambda (estimate beta0 and beta1 via GLM) 
2) with lambda, estimate alpha via ML estimation 
3) with updated alpha, replicate 1) and 2) until alpha is converged to a
value 

My source code is

alpha=rep(1,100)
beta0=rep(0,100)
beta1=rep(0,100)
lambda1=rep(0,100)
lambda2=rep(0,100)
verpi=rep(0,100)
L=rep(0,100)

alpha[0]=1
i=1

while(i<=100){
repeat{

verpi[i]<-c(5^alpha[i-1],10^alpha[i-1]-5^alpha[i-1],14^alpha[i-1]-10^alpha[i-1],18^alpha[i-1]-14^alpha[i-1])
r<-c(1,0,0,1)
k<-c(3,2,2,2)
x<-c(0.5,0.5,1.0,1.0)
GLM_results <- glm(r/k ~ x, family=binomial(link='cloglog'),
offset=log(verpi[i]),weights=k)
beta0[i]<-GLM_results$coefficients[[1]]
beta1[i]<-GLM_results$coefficients[[2]]
lambda1[i]<-beta0[i]+beta1[i]*x[1]
lambda2[i]<-beta0[i]+beta1[i]*x[3]

L[i]<-function(a){
s1_f1<-(exp(-lambda1[i]*(0^a-0^a))-exp(-lambda1[i]*(5^a-0^a)))
s2_f2<-(exp(-lambda1[i]*(10^a)-lambda2[i]*(14^a-10^a+14^a-14^a))
        -exp(-lambda1[i]*(10^a)-lambda2[i]*(14^a-10^a+18^a-14^a)))
logl<-log(s1_f1*s2_f2)
return(-logl)}

optim(1,L[i])
alpha[i]<-round(optim(1,L[i])$par,2)
if(alpha[i]<=alpha[i-1])
break}
i=i+1
}

Best regards,


Jin
-- 
View this message in context:
http://www.nabble.com/Error-in-%22-%3C-%22%28%60*tmp*%60%2C-i%2C-value-%3D-numeric%280%29%29-%3A-tp19464723p19464723.html
Sent from the R help mailing list archive at Nabble.com.

On Fri, 12 Sep 2008, sandsky wrote:
>
> I use "while" loop but it produces an errro. I have no idea about
this.
>
> Error in "[<-"(`*tmp*`, i, value = numeric(0)) :
>        nothing to replace with
What version of R is this?  I suspect not a recent one.  I get

Error in verpi[i] <- c(5^alpha[i - 1], 10^alpha[i - 1] - 5^alpha[i - 1],

and that tells you the problem.  If i = 1, you are looking up alpha[0]. 
(You also try to set it.)  alpha[0] is numeric(0): index '0' means omit 
the value from the result (see any good book on R).
>
> The problem description is
>
> The likelihood includes two parameters to be estimated: lambda
> (=beta0+beta1*x) and alpha. The algorithm for the estimation is as
> following:
> 1) with alpha=0, estimate lambda (estimate beta0 and beta1 via GLM)
> 2) with lambda, estimate alpha via ML estimation
> 3) with updated alpha, replicate 1) and 2) until alpha is converged to a
> value
>
> My source code is
>
> alpha=rep(1,100)
> beta0=rep(0,100)
> beta1=rep(0,100)
> lambda1=rep(0,100)
> lambda2=rep(0,100)
> verpi=rep(0,100)
> L=rep(0,100)
>
> alpha[0]=1
> i=1
>
> while(i<=100){
> repeat{
>
>
verpi[i]<-c(5^alpha[i-1],10^alpha[i-1]-5^alpha[i-1],14^alpha[i-1]-10^alpha[i-1],18^alpha[i-1]-14^alpha[i-1])
> r<-c(1,0,0,1)
> k<-c(3,2,2,2)
> x<-c(0.5,0.5,1.0,1.0)
> GLM_results <- glm(r/k ~ x, family=binomial(link='cloglog'),
> offset=log(verpi[i]),weights=k)
> beta0[i]<-GLM_results$coefficients[[1]]
> beta1[i]<-GLM_results$coefficients[[2]]
> lambda1[i]<-beta0[i]+beta1[i]*x[1]
> lambda2[i]<-beta0[i]+beta1[i]*x[3]
>
> L[i]<-function(a){
> s1_f1<-(exp(-lambda1[i]*(0^a-0^a))-exp(-lambda1[i]*(5^a-0^a)))
> s2_f2<-(exp(-lambda1[i]*(10^a)-lambda2[i]*(14^a-10^a+14^a-14^a))
>        -exp(-lambda1[i]*(10^a)-lambda2[i]*(14^a-10^a+18^a-14^a)))
> logl<-log(s1_f1*s2_f2)
> return(-logl)}
>
> optim(1,L[i])
> alpha[i]<-round(optim(1,L[i])$par,2)
> if(alpha[i]<=alpha[i-1])
> break}
> i=i+1
> }
>
> Best regards,
>
>
> Jin
> -- 
> View this message in context:
http://www.nabble.com/Error-in-%22-%3C-%22%28%60*tmp*%60%2C-i%2C-value-%3D-numeric%280%29%29-%3A-tp19464723p19464723.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595