wilcox_R: What the package does (short line)

adtestl <-
function(x,y,est=tmean,nboot=100,alpha=.05,fr=NA,SEED=TRUE,...){
#
# Test the hypothesis that the regression model is additive.
# Use a variation of Stute et al. (1998, JASA, 93, 141-149).
# method, and running interval version of the backfitting
# algorithm
#
if(!is.matrix(x))stop("X values should be stored in a matrix")
if(ncol(x)==1)stop("There should be two or more predictors")
temp<-cbind(x,y)
p<-ncol(x)
p1<-p+1
temp<-elimna(temp)
x<-temp[,1:p]
x<-as.matrix(x)
y<-temp[,p1]
if(alpha<.05 && nboot<=100)warning("You used alpha<.05 and nboot<=100")
if(is.na(fr)){
fr<-.8
if(ncol(x)==2){
nval<-c(20,30,50,80,150)
fval<-c(0.40,0.36,0.18,0.15,0.09)
if(length(y)<=150)fr<-approx(nval,fval,length(y))$y
if(length(y)>150)fr<-.09
}
}
if(SEED)set.seed(2)
x<-as.matrix(x)
mflag<-matrix(NA,nrow=length(y),ncol=length(y))
for (j in 1:length(y)){
for (k in 1:length(y)){
mflag[j,k]<-(sum(x[j,]<=x[k,])==ncol(x))
}
}
yhat<-adrunl(x,y,plotit=FALSE,fr=fr,pyhat=T)
regres<-y-yhat
print("Taking bootstrap sample, please wait.")
data<-matrix(runif(length(y)*nboot),nrow=nboot)
data<-sqrt(12)*(data-.5) # standardize the random numbers.
rvalb<-apply(data,1,adtestls1,yhat,regres,mflag,x,fr)
# An n x nboot matrix of R values
rvalb<-rvalb/sqrt(length(y))
dstatb<-apply(abs(rvalb),2,max)
wstatb<-apply(rvalb^2,2,mean)
dstatb<-sort(dstatb)
wstatb<-sort(wstatb)
# compute test statistic
v<-c(rep(1,length(y)))
rval<-adtestls1(v,yhat,regres,mflag,x,fr)
rval<-rval/sqrt(length(y))
dstat<-max(abs(rval))
wstat<-mean(rval^2)
ib<-round(nboot*(1-alpha))
critd<-dstatb[ib]
critw<-wstatb[ib]
list(dstat=dstat,wstat=wstat,critd=critd,critw=critw)
}