[R] integrate function - error -integration not occurring with last few rows

Berend Hasselman bhh at xs4all.nl
Fri Apr 6 15:56:00 CEST 2012


On 06-04-2012, at 13:14, Navin Goyal wrote:

> Apologies for the lengthy code.
> I tried a simple (and shorter) piece of code (pasted below) and it still gives me the same error for last few rows. Is this a bug or am I doing something totally wrong?  Could anyone please provide some help/pointers ?
>  

You are not specifying what the  error is for the last few rows?
You are doing something completely wrong (AFAICT).
See below.

> PS.  beta0 was fixed to 0.001 in the previous code. Also if I manually estimated the integral isnt 0. If I scramble the row order, it is again only the last few rows that dont integrate.

See below.

> Thanks
> 
> data1<-expand.grid(1:5,0:6,10)
> names(data1)<-c("ID","TIME", "DOSE")
> data1<-data1[order(data1$DOSE,data1$ID,data1$TIME),]
> ed<-data1[!duplicated(data1$ID) , c(1,3)]
> ed$base=1
> ed$drop=1
> set.seed(5234123)
> k<-0
> for (i in 1:length(ed$ID))
> {
> k<-k+1
> ed$base[k]<-100*exp(rnorm(1,0,0.2))
> ed$drop[k]<-0.20*exp(rnorm(1,0,0.5))
> }

Why are you not using i to index ed$XXX?
You are not using i.
Simplify to

for (k in 1:length(ed$ID))
{
ed$base[k]<-100*exp(rnorm(1,0,0.2))
ed$drop[k]<-0.20*exp(rnorm(1,0,0.5))
}

> comb1<-merge(data1[, c("ID","TIME")], ed)
> comb1$disprog<-comb1$base*exp(-comb1$drop*comb1$TIME)
> comb1$integral=0
> hz.func1<-function(t,bshz,beta1, change)
> { ifelse(t==0,bshz, bshz*exp(beta1*change)) }

Insert here

comb1
i
length(comb1$ID)

and you should see that i is 5 and that length(comb1$ID) is 35.

> q<-0
> for (m in i:length(comb1$ID))
> {
> q<-q+1
> comb1$integral[q]<-integrate(hz.func1, lower=0, upper=comb1$TIME[q],
>           bshz=0.001,beta1=0.035,
>          change=comb1$disprog[q])$value
> }
> comb1
> 

1. Why does your for loop variable m start with i and not 1? (as I told you in my first reply)
2. Why are you not using the for loop variable m?
3. So from the above m starts at 5 and stops at 35 (==> 312 steps)
4, so you are filling elements 1 to 31 of comb1 and items 32 to 35 are unchanged.

5. why don't you do this

for (q in 1:length(comb1$ID))
{
comb1$integral[q]<-integrate(hz.func1, lower=0, upper=comb1$TIME[q],
          bshz=0.001,beta1=0.035,
         change=comb1$disprog[q])$value
}
comb1

This avoids a useless variable m and fill all elements of comb1.
And you could just as well reuse variable k; there is no need for a new variable (q) here.

Berend



More information about the R-help mailing list