I am working on a clinical trial simulation with two groups, treatment and
placebo, and the outcome is dichotomous (recovery or no recovery) . I would
like to stop my loop if either of my conditions are met:
1) futility - there are no responses to treatment in the treatment group.
2) the p-value is significant (<=0.01).
The problem I am having is my loop continues to run 10,000 times even though
I am sure that at least one of the conditions are met in some instances. It
appears the main problem is that my if loop is not adequately filtering the
conditions I specified in it.
library(magrittr)
library(dplyr)
nSims <- 10000 #number of simulated experiments
futility1 <-numeric(nSims) #set up empty container for all simulated
futility
futility2 <-numeric(nSims) #set up empty container for all simulated
futility
significant1 <-numeric(nSims) #set up empty container for all simulated
significance
significant2 <-numeric(nSims) #set up empty container for all simulated
significance
for(i in 1:nSims){ #for each simulated experiment
# Year 1
# p1<-response in controls
# p2<-response in treated
# Generating random deviates from a Uniform(0,1) distribution
control.year1<-(runif(16, min = 0, max = 1))
treat.year1<-(runif(16, min = 0, max = 1))
#Generating dichotomous response variables for each group
control.respond1<-ifelse(control.year1<=0.05,1,0)
treat.respond1<-ifelse(treat.year1<=0.30,1,0)
#Summing number of responses from each group
control.no1<-sum(control.respond1==0)
control.yes1<-sum(control.respond1==1)
treat.no1<-sum(treat.respond1==0)
treat.yes1<-sum(treat.respond1==1)
#Perform the Fisher's exact test (one sided) with p<0.01
fisher<-matrix(c(control.no1,control.yes1,treat.no1,treat.yes1),nrow=2,ncol=2)
f<-fisher.test(fisher,alternative = "greater")
#year 2
if (f$p.value>0.01 && treat.yes1!=0){
# Generating random deviates from a Uniform(0,1) distribution
control.year2<-(runif(16, min = 0, max = 1))
treat.year2<-(runif(16, min = 0, max = 1))
#Generating dichotomous response variables for each group
control.respond2<-ifelse(control.year2<=0.05,1,0)
treat.respond2<-ifelse(treat.year2<=0.30,1,0)
#Summing number of responses from each group
control.no2<-sum(control.respond2==0)
control.yes2<-sum(control.respond2==1)
treat.no2<-sum(treat.respond2==0)
treat.yes2<-sum(treat.respond2==1)
#Perform the Fisher's exact test (one sided) with p<0.01
fisher2<-matrix(c(control.no2,control.yes2,treat.no2,treat.yes2),nrow=2,ncol=2)
f2<-fisher.test(fisher2,alternative = "greater")
}
significant2[i]<-ifelse(f2$p.value<0.01,1,0)
futility2[i]<-ifelse(treat.yes2==0,1,0)
}
table(significant1)
table(futility1)
table(significant2)
table(futility2)
Does this example help?> for (ii in 1:10) { cat( ii,'\n') ; if (ii >3) break }1 2 3 4 The loop won't stop unless you tell it to stop, and I don't see any place where you told it to stop. -- Don MacQueen Lawrence Livermore National Laboratory 7000 East Ave., L-627 Livermore, CA 94550 925-423-1062 Lab cell 925-724-7509 ?On 6/28/18, 12:53 PM, "R-help on behalf of Kelly Wu" <r-help-bounces at r-project.org on behalf of kekwu at ucdavis.edu> wrote: I am working on a clinical trial simulation with two groups, treatment and placebo, and the outcome is dichotomous (recovery or no recovery) . I would like to stop my loop if either of my conditions are met: 1) futility - there are no responses to treatment in the treatment group. 2) the p-value is significant (<=0.01). The problem I am having is my loop continues to run 10,000 times even though I am sure that at least one of the conditions are met in some instances. It appears the main problem is that my if loop is not adequately filtering the conditions I specified in it. library(magrittr) library(dplyr) nSims <- 10000 #number of simulated experiments futility1 <-numeric(nSims) #set up empty container for all simulated futility futility2 <-numeric(nSims) #set up empty container for all simulated futility significant1 <-numeric(nSims) #set up empty container for all simulated significance significant2 <-numeric(nSims) #set up empty container for all simulated significance for(i in 1:nSims){ #for each simulated experiment # Year 1 # p1<-response in controls # p2<-response in treated # Generating random deviates from a Uniform(0,1) distribution control.year1<-(runif(16, min = 0, max = 1)) treat.year1<-(runif(16, min = 0, max = 1)) #Generating dichotomous response variables for each group control.respond1<-ifelse(control.year1<=0.05,1,0) treat.respond1<-ifelse(treat.year1<=0.30,1,0) #Summing number of responses from each group control.no1<-sum(control.respond1==0) control.yes1<-sum(control.respond1==1) treat.no1<-sum(treat.respond1==0) treat.yes1<-sum(treat.respond1==1) #Perform the Fisher's exact test (one sided) with p<0.01 fisher<-matrix(c(control.no1,control.yes1,treat.no1,treat.yes1),nrow=2,ncol=2) f<-fisher.test(fisher,alternative = "greater") #year 2 if (f$p.value>0.01 && treat.yes1!=0){ # Generating random deviates from a Uniform(0,1) distribution control.year2<-(runif(16, min = 0, max = 1)) treat.year2<-(runif(16, min = 0, max = 1)) #Generating dichotomous response variables for each group control.respond2<-ifelse(control.year2<=0.05,1,0) treat.respond2<-ifelse(treat.year2<=0.30,1,0) #Summing number of responses from each group control.no2<-sum(control.respond2==0) control.yes2<-sum(control.respond2==1) treat.no2<-sum(treat.respond2==0) treat.yes2<-sum(treat.respond2==1) #Perform the Fisher's exact test (one sided) with p<0.01 fisher2<-matrix(c(control.no2,control.yes2,treat.no2,treat.yes2),nrow=2,ncol=2) f2<-fisher.test(fisher2,alternative = "greater") } significant2[i]<-ifelse(f2$p.value<0.01,1,0) futility2[i]<-ifelse(treat.yes2==0,1,0) } table(significant1) table(futility1) table(significant2) table(futility2) ______________________________________________ R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see 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.
Hi Kelly,
You seem to be testing the two "years" independently, so here is a
possible solution:
npg<-16
futility1<-futility2<-psignif1<-psignif2<-nrep<-0
while(!futility1 && !futility2 && !psignif1 && !psignif2
&& nrep < 10000) {
control_respond1<-sum(runif(npg,0,1) < 0.05)
control_respond2<-sum(runif(npg,0,1) < 0.05)
treat_respond1<-sum(runif(npg,0,1) < 0.3)
treat_respond2<-sum(runif(npg,0,1) < 0.3)
futility1<-treat_respond1 == 0
futility2<-treat_respond2 == 0
psignif1<-fisher.test(matrix(c(control_respond1,npg-control_respond1,
treat_respond1,npg-treat_respond1),nrow=2),
alternative="greater")$p.value < 0.01
psignif2<-fisher.test(matrix(c(control_respond2,npg-control_respond2,
treat_respond2,npg-treat_respond2),nrow=2),
alternative="greater")$p.value < 0.01
nrep<-nrep+1
}
cat("futility1",futility1,"futility2",futility2,"psignif1",psignif1,
"psignif2",psignif2,"nrep",nrep,"\n")
The output tells you which condition was met and on which repetition
it occurred. The outcomes for all previous repetitions will be FALSE
and for the remaining repetitions, undefined (unless you have a cat
named Schrodinger)..
If you want to test the two "years" as sequential observations, it
will only require a minor modification.
Jim
On Fri, Jun 29, 2018 at 5:53 AM, Kelly Wu <kekwu at ucdavis.edu>
wrote:> I am working on a clinical trial simulation with two groups, treatment and
> placebo, and the outcome is dichotomous (recovery or no recovery) . I would
> like to stop my loop if either of my conditions are met:
>
> 1) futility - there are no responses to treatment in the treatment group.
> 2) the p-value is significant (<=0.01).
>
> The problem I am having is my loop continues to run 10,000 times even
though
> I am sure that at least one of the conditions are met in some instances. It
> appears the main problem is that my if loop is not adequately filtering the
> conditions I specified in it.
>
> library(magrittr)
> library(dplyr)
>
> nSims <- 10000 #number of simulated experiments
> futility1 <-numeric(nSims) #set up empty container for all simulated
> futility
> futility2 <-numeric(nSims) #set up empty container for all simulated
> futility
>
> significant1 <-numeric(nSims) #set up empty container for all simulated
> significance
> significant2 <-numeric(nSims) #set up empty container for all simulated
> significance
>
> for(i in 1:nSims){ #for each simulated experiment
> # Year 1
>
> # p1<-response in controls
> # p2<-response in treated
> # Generating random deviates from a Uniform(0,1) distribution
> control.year1<-(runif(16, min = 0, max = 1))
> treat.year1<-(runif(16, min = 0, max = 1))
>
> #Generating dichotomous response variables for each group
> control.respond1<-ifelse(control.year1<=0.05,1,0)
> treat.respond1<-ifelse(treat.year1<=0.30,1,0)
>
> #Summing number of responses from each group
> control.no1<-sum(control.respond1==0)
> control.yes1<-sum(control.respond1==1)
> treat.no1<-sum(treat.respond1==0)
> treat.yes1<-sum(treat.respond1==1)
>
> #Perform the Fisher's exact test (one sided) with p<0.01
>
>
fisher<-matrix(c(control.no1,control.yes1,treat.no1,treat.yes1),nrow=2,ncol=2)
> f<-fisher.test(fisher,alternative = "greater")
>
> #year 2
> if (f$p.value>0.01 && treat.yes1!=0){
> # Generating random deviates from a Uniform(0,1) distribution
> control.year2<-(runif(16, min = 0, max = 1))
> treat.year2<-(runif(16, min = 0, max = 1))
>
> #Generating dichotomous response variables for each group
> control.respond2<-ifelse(control.year2<=0.05,1,0)
> treat.respond2<-ifelse(treat.year2<=0.30,1,0)
>
> #Summing number of responses from each group
> control.no2<-sum(control.respond2==0)
> control.yes2<-sum(control.respond2==1)
> treat.no2<-sum(treat.respond2==0)
> treat.yes2<-sum(treat.respond2==1)
>
> #Perform the Fisher's exact test (one sided) with p<0.01
>
>
fisher2<-matrix(c(control.no2,control.yes2,treat.no2,treat.yes2),nrow=2,ncol=2)
> f2<-fisher.test(fisher2,alternative = "greater")
> }
>
>
> significant2[i]<-ifelse(f2$p.value<0.01,1,0)
> futility2[i]<-ifelse(treat.yes2==0,1,0)
> }
>
> table(significant1)
> table(futility1)
>
> table(significant2)
> table(futility2)
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.
Hi Jim & Don, I was trying to use the break command originally before posting but for some reason it was making almost all of the p-values in the replications non significant. I think I am going to change the flow of the loop so I don?t have to use a break, such as the code Jim wrote. Thanks for your detailed response! Kelly> On Jun 28, 2018, at 3:18 PM, Jim Lemon <drjimlemon at gmail.com> wrote: > > " as sequential observations, it > will only require a minor modification. > > Jim > > On Fri, Jun 29, 2018 at 5:53 AM, Kelly Wu <kekwu at ucdavis.edu> wrote: >[[alternative HTML version deleted]]