library(boot)
library(MASS)
library(base)
library(stats)

corr=function(y,x)
{
	a=var(y)*var(x)
	if(a==0) co=0
	else co=cov(y,x)/sqrt(a)
	co
}


##################################################################
##	AW1 is an association test to test association		##
##	between a phenotype and a group of variants (can	##
##	be rare and common) in a genomic region by using an	##
##	adaptive weighting method. AW1 is designed for the	##
##	case of no or small proportion of protective variants.	##
##	y represents trait values(a vector). x represents 	##
##	genotype (number of minor allele) matrix. Each row	## 
##	represents an individual and each column represents	##
##	a SNP. 							##
##################################################################
AW1=function(y,x,numper)
{
	n=length(y)
	Tper=rep(0,numper)
	if(n==length(x)) 
	{
		T=corr(y,x)
		T=T*(T>0)
		for (i in 1:numper)
		{
			yper=sample(y)
			Tper[i]=corr(yper,x)
			Tper[i]=Tper[i]*(Tper[i]>0)
		}		
	}
	################################
	else
	{
	m=ncol(x)
	q=(colSums(x)+1)/(n*2+2)
	w=1/sqrt(q*(1-q))
	ww=cor(y,x)
	w1=as.numeric(w*ww*(ww>0))
	x1=t(w1*t(x))
	x1=rowSums(x1)
	T=corr(y,x1)
	for(i in 1:numper)
		{
		yper=sample(y)
		ww=cor(yper,x)
		w1=as.numeric(w*ww*(ww>0))
		x1=t(w1*t(x))
		x1=rowSums(x1)
		Tper[i]=corr(yper,x1)
		}
	}
pvalue=sum(Tper>T)/numper
}

##################################################################
##	AW2 is an association test to test association		##
##	between a phenotype and a group of variants (can	##
##	be rare and common) in a genomic region by using an	##
##	adaptive weighting method. AW2 is designed for the	##
##	case when both risk and protective variants are 	##
##	presented.						##
##	y represents trait values(a vector). x represents 	##
##	genotype (number of minor allele) matrix. Each row	## 
##	represents an individual and each column represents	##
##	a SNP. 							##
##################################################################

AW2=function(y,x,numper)
{
	n=length(y)
	Tper=rep(0,numper)
	if(n==length(x)) 
	{
		T=corr(y,x)
		T=T^2
		for (i in 1:numper)
		{
			yper=sample(y)
			Tper[i]=corr(yper,x)
			Tper[i]=Tper[i]^2
		}		
	}
	################################
	else
	{
	m=ncol(x)
	q=(colSums(x)+1)/(n*2+2)
	w=1/sqrt(q*(1-q))
	ww=cor(y,x)
	w1=as.numeric(w*ww)
	x1=t(w1*t(x))
	x1=rowSums(x1)
	T=corr(y,x1)
	for(i in 1:numper)
		{
		yper=sample(y)
		ww=cor(yper,x)
		w1=as.numeric(w*ww)
		x1=t(w1*t(x))
		x1=rowSums(x1)
		Tper[i]=corr(yper,x1)
		}
	}
pvalue=sum(Tper>T)/numper
}