Dear R-users,
I'm having a small problem while bootstraping data.
What i would like to do, is resmple the data and calulate a function on
this, so i can estimate the measure of reproducability for this data.
The function i wrote works fine, even while bootstraping.
The only problem is that bootstraping.
The dataset existes of 10 trials, each divided in to 3 groups of high(3)
medium(2) and low(1).
A bootstrap samlpe (trial) should always exist of 5 obs. taken from each
group population, so to be representative.
example:
original data:
trial 1 : group(1) = (0,0,1,0,0);group(2) = (0,1,1,0,1);group(3) =
(1,1,1,1,1)
...
bootstraped data:
trial 1 : group(1) = (0,0,0,0,1);group(2) = (1,1,0,0,1);group(3) =
(1,0,1,1,1)
NOT
bootstraped data:
trial 1 : group(1) = (0,0,0,0,1,1,0);group(2) = (1,0,1);group(3) =
(1,0,1,1,1,1,1,0,1,1)
Now I am familiar how to use function "bootstrap" (pkg "bootstrap"), but i
read about a function called "boot" (pkg "boot"), however i can't seem to
master this.
The explanation (help('boot') ) isn't making me any smarter.
I know I can always split the data up (wich is what i am doing) but i was
wondering whether this would have an effect on the bootstrap, maby it is
beter to keep all the groups together?
here is a (this time WORKING) code example of what i did.
## proc
## generate data
datas <-
data.frame("protection"=c(rep(c(0,1,0,1,0,0,1,0,1,1,1,0,1,1,1),2),c(0,0,0,0,1,0,1,0,1,1,1,0,1,0,1),rep(c(0,1,1,1,0,0,1,1,0,1,1,1,1,0,1),2),c(0,1,0,0,1,0,1,1,1,1,1,1,1,1,1),rep(c(0,1,0,1,0,0,1,1,1,1,1,0,1,1,1),2),c(0,1,0,0,0,0,1,1,0,1,1,1,1,1,1),c(0,0,1,1,0,0,1,1,1,0,0,1,1,0,1))
,"group"=rep(1:3,50),"trial"=c(rep(1,15),rep(2,15),rep(3,15),rep(4,15),rep(5,15),rep(6,15),rep(7,15),rep(8,15),rep(9,15),rep(10,15)))
## describe Function
Vacc.Vcon <-function (dataset1 , trialdata , groupdata ) {
groups <- unique (groupdata)
trials <- unique (trialdata)
Tr <- length(trials)
G <- length(groups)
Gl <- length(dataset1)/(G*Tr)
Tl <- length(dataset1)/(Tr)
iterg <-data.frame(1:G,as.vector(groups))
trials <- unique (trialdata)
Tr <- length(trials)
itert <-data.frame(1:Tr,as.vector(trials))
triallist <- c()
grouplist <- c()
for (x in 1:G){
ifelse(x==1,y<-x,y<- y+Tr)
grouplist[c(y:(y+Tr-1))] <-rep(iterg[x,2],Tr)}
iter <-data.frame(1:(Tr),rep(trials,G),grouplist)
VACC <- data.frame()
VACC.sub <- function (dataset1,trialn,groupn){
p0 <-sum( ifelse(dataset1==1 & trialdata==trialn &
groupdata==groupn, 1,0) )
p1 <-sum( ifelse(dataset1==0 & trialdata==trialn &
groupdata==groupn, 1,0) )
p <- p0+p1
VACC.group <-
list('Trial'=trialn,'Group'=groupn,'Vacc'=sum((p0/p)^2
,(p1/p)^2),"p0"=(p0/p) , "p1"=(p1/p)
,"n0"=as.numeric(p0),"n1"=as.numeric(p1),'n'=as.numeric(p))
VACC.group }
for (i in 1:(G*Tr) ) {
VACC[i,1] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[1]
VACC[i,2] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[2]
VACC[i,3] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[3]
VACC[i,4] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[4]
VACC[i,5] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[5]
VACC[i,6] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[6]
VACC[i,7] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[7]
VACC[i,8] <- VACC.sub (dataset1,iter[i,2],iter[i,3])[8]
VACC}
rownames(VACC) <- NULL
rownames(VACC) <- paste(iter[,2],iter[,3],sep='_')
Pcalc <- function(x) {
out<-(1/(Tr)) * sum(x)
out}
P0 <- tapply( VACC$p0,VACC$Group,Pcalc)
P1 <- tapply( VACC$p1,VACC$Group,Pcalc)
Vcon <- mean(cbind(P0^2 + P1^2))
Vacc.total <- mean (tapply( VACC$Vacc,VACC$Group,mean))
out <-
list("all"=VACC,"N"=G,"P0"=P0,"P1"=P1,"Vcon"=Vcon*100,"Vacc.total"=Vacc.total*100)
out }
## end describe Function
Vacc.Vcon (datas[,1] , datas[,3], datas[,2]) # example of how fun works
## data needs to be in matrix form for bootstrap function
xdata <-matrix(
cbind(datas$protection,datas$group,datas$trial),ncol=3,byrow=F)
## function for bootstrap
vacc.boot <- function(x,xdata){
Vacc.Vcon(xdata[x,1],xdata[x,3],xdata[x,2]) }
bootk <- 10
results <- bootstrap(1:150,bootk,vacc.boot,xdata)
taccs <- list() ;Vaccs <- vector();Vcons <- vector()
boot.amp.vac2<- for(i in 1:bootk) {
m.i <- results$thetastar[[i]]
taccs[i] <- list(m.i )
G.Vacc <- round(
tapply(taccs[[i]]$all$Vacc,rownames(taccs[[i]]$all),mean)*100 ,digits=3)
Vaccs <- round( mean(taccs[[i]]$Vacc.total),digits=3)
Vcons<-round( mean(taccs[[i]]$Vcon ),digits=3)
tacc <- list(
"data"=taccs,"Booted.means"=list("Vacc.grouped"=G.Vacc
,"Vacc.Total"=Vaccs,"Vcon.Total"=Vcons))
tacc}
Rep.table <- tacc$Booted.mean
Rep.table
## problem area => n should always be 5 in each group as in the original
data
#calcues based on original data last colon : n = 5
Vacc.Vcon (datas[,1] ,datas[,3], datas[,2])$all [1:5,]
#calcues based on Booted data n is not 5 !
tacc$data[[2]]$all[1:5,]
tacc$data[[10]]$all[1:5,]
## does not work at all
# i thoughed that f means the strata have to be used
# as frequenties, for resampling from this (strata)group
boot(xdata, Vacc.Vcon, R=bootk, stype="f", strata=xdata[,3])
boot(xdata, vacc.boot, R=bootk, stype="f", strata=xdata[,3])
## end proc
Thanks in advance,
Tom.
Disclaimer: click here
[[alternative HTML version deleted]]
______________________________________________
R-help@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.
