[R] Estimating parameters of 2 phase Coxian using optim

[Andy Fugard]

Hi There,

Perhaps the problem is the line

         loglik<-log(p %*% expm(Q * y(i)) %*% q)

You mention that y is a vector but here you're treating it as a  
function.  Maybe try

         loglik<-log(p %*% expm(Q * y[i]) %*% q)

?

Don't have a clue about the correctness of the contents of cox2.lik...

Andy


On 6 Mar 2007, at 08:54, Laura Hill wrote:

> Hi,
>
> My name is Laura. I'm a PhD student at Queen's University Belfast  
> and have
> just started learning R. I was wondering if somebody could help me  
> to see
> where I am going wrong in my code for estimating the parameters  
> [mu1, mu2,
> lambda1] of a 2-phase Coxian Distribution.
>
> cox2.lik<-function(theta, y){
>     mu1<-theta[1]
>
>     mu2<-theta[2]
>
>     lambda1<-theta[3]
>
>     p<-Matrix(c(1, 0), nrow=1, ncol=2)
>
>     Q<-Matrix(c(-(lambda1 + mu1), 0, lambda1, -mu2), nrow=2, ncol=2)
>
>     q<-Matrix(c(mu1, mu2), nrow=2, ncol=1)
>
>     for (i in 1:length(y)){
>         loglik<-log(p %*% expm(Q * y(i)) %*% q)
>     return(loglik)}
>
>     sumloglik<-sum(loglik)
>
>     return(-sumloglik)
>     }
>
> I have installed the Matrix package. y is a vector of 240 survival  
> times. In
> the R console I typed
>
>> source("/private/var/automount/users/lhill07/Desktop/cox2.lik.R")
>
>> optim(c(0.5, 0.5, 0.5), cox2.lik, y=y, method="BFGS")
>
> Error: could not find function "y"
>
> Error in expm(Q * y(i)) : error in evaluating the argument 'x' in  
> selecting
> a method for function 'expm'
>
> Error in log(p %*% expm(Q * y(i)) %*% q) :
>     error in evaluating the argument 'x' in selecting a method for  
> function
> 'log'
>
>
> I'm sorry if I have missed something really obvious and I would  
> appreciate
> any help that is offered. Hopefully I have given enough information.
>
> Many thanks in advance,
>
> Laura
>
> ______________________________________________
> R-help at stat.math.ethz.ch 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.


--
Andy Fugard, Postgraduate Research Student
Psychology (Room F15), The University of Edinburgh,
   7 George Square, Edinburgh EH8 9JZ, UK
Mobile: +44 (0)78 123 87190   http://www.possibly.me.uk