wilcox_R: What the package does (short line)

ancboot <-
function(x1,y1,x2,y2,fr1=1,fr2=1,tr=.2,nboot=599,pts=NA,plotit=TRUE,xout=FALSE,outfun=outpro,...){
#
# Compare two independent  groups using the ancova method
# in chapter 11 of Wilcox, 2013, Intro to Robust Estimation and Hypothesis Testing. 
# No assumption is made about the form of the regression
# lines--a running interval smoother is used.
# Confidence intervals are computed using a percentile t bootstrap
# method. Comparisons are made at five empirically chosen design points.
#
#  Assume data are in x1 y1 x2 and y2
#
if(is.na(pts[1])){
isub<-c(1:5)  # Initialize isub
test<-c(1:5)
m1=elimna(cbind(x1,y1))
x1=m1[,1]
y1=m1[,2]
m1=elimna(cbind(x2,y2))
x2=m1[,1]
y2=m1[,2]
xorder<-order(x1)
y1<-y1[xorder]
x1<-x1[xorder]
xorder<-order(x2)
y2<-y2[xorder]
x2<-x2[xorder]
n1<-1
n2<-1
vecn<-1
for(i in 1:length(x1))n1[i]<-length(y1[near(x1,x1[i],fr1)])
for(i in 1:length(x1))n2[i]<-length(y2[near(x2,x1[i],fr2)])
for(i in 1:length(x1))vecn[i]<-min(n1[i],n2[i])
sub<-c(1:length(x1))
isub[1]<-min(sub[vecn>=12])
isub[5]<-max(sub[vecn>=12])
isub[3]<-floor((isub[1]+isub[5])/2)
isub[2]<-floor((isub[1]+isub[3])/2)
isub[4]<-floor((isub[3]+isub[5])/2)
mat<-matrix(NA,5,8)
dimnames(mat)<-list(NULL,c("X","n1","n2","DIF","TEST","ci.low","ci.hi",
"p.value"))
gv1<-vector("list")
for (i in 1:5){
j<-i+5
temp1<-y1[near(x1,x1[isub[i]],fr1)]
temp2<-y2[near(x2,x1[isub[i]],fr2)]
temp1<-temp1[!is.na(temp1)]
temp2<-temp2[!is.na(temp2)]
mat[i,2]<-length(temp1)
mat[i,3]<-length(temp2)
gv1[[i]]<-temp1
gv1[[j]]<-temp2
}
I1<-diag(5)
I2<-0-I1
con<-rbind(I1,I2)
test<-linconb(gv1,con=con,tr=tr,nboot=nboot)
for(i in 1:5){
mat[i,1]<-x1[isub[i]]
}
mat[,4]<-test$psihat[,2]
mat[,5]<-test$test[,2]
mat[,6]<-test$psihat[,3]
mat[,7]<-test$psihat[,4]
mat[,8]<-test$test[,4]
}
if(!is.na(pts[1])){
n1<-1
n2<-1
vecn<-1
for(i in 1:length(pts)){
n1[i]<-length(y1[near(x1,pts[i],fr1)])
n2[i]<-length(y2[near(x2,pts[i],fr2)])
if(n1[i]<=5)paste("Warning, there are",n1[i]," points corresponding to the design point X=",pts[i])
if(n2[i]<=5)paste("Warning, there are",n2[i]," points corresponding to the design point X=",pts[i])
}
mat<-matrix(NA,length(pts),9)
dimnames(mat)<-list(NULL,c("X","n1","n2","DIF","TEST","se","ci.low","ci.hi",
"p.value"))
gv<-vector("list",2*length(pts))
for (i in 1:length(pts)){
g1<-y1[near(x1,pts[i],fr1)]
g2<-y2[near(x2,pts[i],fr2)]
g1<-g1[!is.na(g1)]
g2<-g2[!is.na(g2)]
j<-i+length(pts)
gv[[i]]<-g1
gv[[j]]<-g2
}
I1<-diag(length(pts))
I2<-0-I1
con<-rbind(I1,I2)
test<-linconb(gv,con=con,tr=tr,nboot=nboot)
mat[,1]<-pts
mat[,2]<-n1
mat[,3]<-n2
mat[,4]<-test$psihat[,2]
mat[,5]<-test$test[,2]
mat[,6]<-test$test[,3]
mat[,7]<-test$psihat[,3]
mat[,8]<-test$psihat[,4]
mat[,9]<-test$test[,4]
}
if(plotit){
if(xout){
flag<-outfun(x1,...)$keep
x1<-x1[flag]
y1<-y1[flag]
flag<-outfun(x2,...)$keep
x2<-x2[flag]
y2<-y2[flag]
}
runmean2g(x1,y1,x2,y2,fr=fr1,est=mean,tr=tr)
}
list(output=mat,crit=test$crit)
}