R help - Oct 2012 - Error in integrate(integrand, 0, Inf) : non-finite function value

Dear R users,

When I run the code below, I get the error "Error in integrate(integrand,
0,
Inf) : non-finite function value".  The code works if the function returns
only "sum(integ)". However, I want to add "cmh" to it. When
I add "cmh" I
get that error. I can't figure out why this is happening because my
integrate function has nothing to do with "cmh".  I tried to integrate
from
0 to 1000, and still same error. Any suggestion is greatly appreciated.
Thank you in advance!



d<-matrix(c(1,1,0,0,0,0,0,0,2,1,0,0,1,1,0,1,2,2,1,0),nrow=10,ncol=2)
h<-matrix(runif(20,0,1),10)
delta<-matrix(c(2,1,0,1,0,1,0,0,2,1,0,0,1,1,1,1,0,2,1,0),nrow=10,ncol=2)

out<-vector("numeric",length(1:2))
integ<-vector("numeric",length(1:2))

for (k in 1:2){
ll<-function(p){
cmh<-delta[,k]*(h[,k]*log(0.5))+p*log(gamma(1+1/p))
for(s in 1:10){
integrand<-function(x)
x^d[s,k]*exp(-x*gamma(1+1/p))^p*p*x^(p-1)*exp(-x*h[s,k]) 
}
integ<-integrate(integrand,0,Inf)$value
cmhn<-as.vector(cmh)
lik<-sum(cmhn+integ)
-lik
}
initial<-c(1)
t<-nlm(ll,initial)
out[k]<-t$estimate
}

est<-as.vector(out)



--
View this message in context:
http://r.789695.n4.nabble.com/Error-in-integrate-integrand-0-Inf-non-finite-function-value-tp4646868.html
Sent from the R help mailing list archive at Nabble.com.

On Oct 20, 2012, at 9:23 AM, stats12 wrote:
> Dear R users,
> 
> When I run the code below, I get the error "Error in
integrate(integrand, 0,
> Inf) : non-finite function value".  The code works if the function
returns
> only "sum(integ)".
But you never showed us the working code.
> However, I want to add "cmh" to it. When I add "cmh" I
> get that error. I can't figure out why this is happening because my
> integrate function has nothing to do with "cmh".  I tried to
integrate from
> 0 to 1000, and still same error. Any suggestion is greatly appreciated.
> Thank you in advance!
> 
> 
> 
> d<-matrix(c(1,1,0,0,0,0,0,0,2,1,0,0,1,1,0,1,2,2,1,0),nrow=10,ncol=2)
> h<-matrix(runif(20,0,1),10)
> delta<-matrix(c(2,1,0,1,0,1,0,0,2,1,0,0,1,1,1,1,0,2,1,0),nrow=10,ncol=2)
> 
> out<-vector("numeric",length(1:2))
> integ<-vector("numeric",length(1:2))
> 
> for (k in 1:2){
> ll<-function(p){
>      cmh<-delta[,k]*(h[,k]*log(0.5))+p*log(gamma(1+1/p))
This inner loop appears to define a function 10 times, but does nothing else:
>      for(s in 1:10){
>      integrand<-function(x)
>      x^d[s,k]*exp(-x*gamma(1+1/p))^p*p*x^(p-1)*exp(-x*h[s,k]) 
>                     }  # end of s-loop
> ------------------
>      integ<-integrate(integrand,0,Inf)$value
>      cmhn<-as.vector(cmh)
>      lik<-sum(cmhn+integ)
>      -lik
>               }  # end of ll()-function
> initial<-c(1)
> t<-nlm(ll,initial)
> out[k]<-t$estimate
> } # end of k-loop
> est<-as.vector(out)

Your uncommented code seems to have problems with organization, but since you
never really described your overall strategy or goal was, it's hard to know.

-- 


David Winsemius, MD
Alameda, CA, USA

Hi,

Thank you for your comment. I worked on the code again and was able to make
it work. The only problem I am having right now is that nlm depends on the
initial value.  

When the initial value is 1, I get the following estimates
0.1230414 19.6271029

when it is 2, I get the following
29.46874 20.01679



d<-matrix(c(1,1,0,0,0,0,0,0,2,1,0,0,1,1,0,1,2,2,1,0),nrow=10,ncol=2) 
h<-matrix(runif(20,0,1),10) 
delta<-matrix(c(2,1,0,1,0,1,0,0,2,1,0,0,1,1,1,1,0,2,1,0),nrow=10,ncol=2) 

out<-vector("numeric",length(1:2)) 
integ<-vector("numeric",length(1:10)) 

for (k in 1:2){ 
ll<-function(p){ 
cmh<-delta[,k]*(h[,k]*log(0.5))+p*log(gamma(1+1/p)) 
for(s in 1:10){ 
integrand<-function(x)
x^d[s,k]*exp(-x*gamma(1+1/p))^p*p*x^(p-1)*exp(-x*h[s,k]) 
integ<-integrate(integrand,0,Inf)$value 
return(integ)
}

lik<-sum(cmh+log(integ)) 
-lik 
} 
initial<-c(1) 
t<-nlm(ll,initial) 
out[k]<-t$estimate 
} 

est<-as.vector(out)








--
View this message in context:
http://r.789695.n4.nabble.com/Error-in-integrate-integrand-0-Inf-non-finite-function-value-tp4646868p4646896.html
Sent from the R help mailing list archive at Nabble.com.

That is an intrinsic part of nonlinear optimization. Choose your starting point
wisely.
---------------------------------------------------------------------------
Jeff Newmiller                        The     .....       .....  Go Live...
DCN:<jdnewmil at dcn.davis.ca.us>        Basics: ##.#.       ##.#.  Live
Go...
                                      Live:   OO#.. Dead: OO#..  Playing
Research Engineer (Solar/Batteries            O.O#.       #.O#.  with
/Software/Embedded Controllers)               .OO#.       .OO#.  rocks...1k
--------------------------------------------------------------------------- 
Sent from my phone. Please excuse my brevity.

stats12 <skarmv at gmail.com> wrote:
>Hi,
>
>Thank you for your comment. I worked on the code again and was able to
>make
>it work. The only problem I am having right now is that nlm depends on
>the
>initial value.  
>
>When the initial value is 1, I get the following estimates
>0.1230414 19.6271029
>
>when it is 2, I get the following
>29.46874 20.01679
>
>
>
>d<-matrix(c(1,1,0,0,0,0,0,0,2,1,0,0,1,1,0,1,2,2,1,0),nrow=10,ncol=2) 
>h<-matrix(runif(20,0,1),10) 
>delta<-matrix(c(2,1,0,1,0,1,0,0,2,1,0,0,1,1,1,1,0,2,1,0),nrow=10,ncol=2)
>
>
>out<-vector("numeric",length(1:2)) 
>integ<-vector("numeric",length(1:10)) 
>
>for (k in 1:2){ 
>ll<-function(p){ 
>cmh<-delta[,k]*(h[,k]*log(0.5))+p*log(gamma(1+1/p)) 
>for(s in 1:10){ 
>integrand<-function(x)
>x^d[s,k]*exp(-x*gamma(1+1/p))^p*p*x^(p-1)*exp(-x*h[s,k]) 
>integ<-integrate(integrand,0,Inf)$value 
>return(integ)
>}
>
>lik<-sum(cmh+log(integ)) 
>-lik 
>} 
>initial<-c(1) 
>t<-nlm(ll,initial) 
>out[k]<-t$estimate 
>} 
>
>est<-as.vector(out)
>
>
>
>
>
>
>
>
>--
>View this message in context:
>http://r.789695.n4.nabble.com/Error-in-integrate-integrand-0-Inf-non-finite-function-value-tp4646868p4646896.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.

On 21-10-2012, at 05:01, stats12 wrote:
> Hi,
> 
> Thank you for your comment. I worked on the code again and was able to make
> it work. The only problem I am having right now is that nlm depends on the
> initial value.  
> 
> When the initial value is 1, I get the following estimates
> 0.1230414 19.6271029
> 
> when it is 2, I get the following
> 29.46874 20.01679
> 
> 
> 
> d<-matrix(c(1,1,0,0,0,0,0,0,2,1,0,0,1,1,0,1,2,2,1,0),nrow=10,ncol=2) 
> h<-matrix(runif(20,0,1),10) 
> delta<-matrix(c(2,1,0,1,0,1,0,0,2,1,0,0,1,1,1,1,0,2,1,0),nrow=10,ncol=2)
> 
> out<-vector("numeric",length(1:2)) 
> integ<-vector("numeric",length(1:10)) 
> 
> for (k in 1:2){ 
> ll<-function(p){ 
> cmh<-delta[,k]*(h[,k]*log(0.5))+p*log(gamma(1+1/p)) 
> for(s in 1:10){ 
> integrand<-function(x)
> x^d[s,k]*exp(-x*gamma(1+1/p))^p*p*x^(p-1)*exp(-x*h[s,k]) 
> integ<-integrate(integrand,0,Inf)$value 
> return(integ)
> }
> 
> lik<-sum(cmh+log(integ)) 
> -lik 
> } 
> initial<-c(1) 
> t<-nlm(ll,initial) 
> out[k]<-t$estimate 
> } 
> 
> est<-as.vector(out)
When I run this code (in TextMate 1.5; Mac OS X 10.6.8)  I get:

Error in integrate(integrand, 0, Inf) : non-finite function value
RMate stopped at line 0
Calls: nlm -> <Anonymous> -> f -> integrate
In addition: Warning messages:
1: In log(gamma(1 + 1/p)) : NaNs produced
2: In log(gamma(1 + 1/p)) : NaNs produced
Execution halted

In addition: why return(integ) in the inner loop with s? This implies an
immediate return in function ll.

You initialize integ as a vector. But in the inner s loop you assign a scalar to
integ.
The whole thing looks very muddled.

Your code should be indented for clarity.

Berend