wilcox_R: What the package does (short line)

qrchkv2 <-
function(x,y,qval=.5,...){
#
# Test of a linear fit based on quantile regression
# The method stems from He and Zhu 2003, JASA, 98, 1013-1022.
# Here, resampling is avoided using approximate critical values if
# com.pval=F
#
#  To get a p-value, via simulations, set  com.pval=T
#  nboot is number of simulations used to determine p-value.
#  Execution time can be quite high
#
#  This function quickly determines .1, .05, .025 and .01
#  critical values for n<=400 and p<=6 (p= number of predictors)
#  and when dealing with the .5 quantile.
#  Otherwise, critical values are determined via simulations, which
#  can have high execution time.
#
#  But, once critical values are determined for a given n, p and
#  quantile qval, the function will remember these values and use them
# in the future. They are stored in a file called qrchk.crit
# Currently, however, when you source the Rallfun files, these values
#  will be lost. You might save the file qrchk.crit in another file,
# source Rallfun, then copy the save file back to qrchk.crit
#
x=as.matrix(x)
p<-ncol(x)
pp1<-p+1
yx<-elimna(cbind(y,x)) #Eliminate missing values.
y<-yx[,1]
x<-yx[,2:pp1]
store.it=F
x<-as.matrix(x)
p.val<-NULL
crit.val<-NULL
x<-as.matrix(x)
# shift the  marginal x values so that the test statistic is
# invariant under changes in location
n<-length(y)
x=standm(x)
gdot<-cbind(rep(1,n),x)
gdot<-ortho(gdot)
x<-gdot[,2:pp1]
x<-as.matrix(x)
temp<-rqfit(x,y,qval=qval,res=T)
coef<-temp$coef
psi<-NA
psi<-ifelse(temp$residuals>0,qval,qval-1)
rnmat<-matrix(0,nrow=n,ncol=pp1)
ran.mat<-apply(x,2,rank)
flagvec<-apply(ran.mat,1,max)
for(j in 1:n){
flag<-ifelse(flagvec<=flagvec[j],T,F)
flag<-as.numeric(flag)
rnmat[j,]<-apply(flag*psi*gdot,2,sum)
}
rnmat<-rnmat/sqrt(n)
temp<-matrix(0,pp1,pp1)
for(i in 1:n)temp<-temp+rnmat[i,]%*%t(rnmat[i,])
temp<-temp/n
test<-max(eigen(temp)$values)
test
}