load("/framshare/fulldata.Rdata")
fulldata$female-1->fulldata$female
paste(fulldata$idmoth,fulldata$idfath,sep="_")->fulldata$family_id
fulldata[fulldata$generation==3,]->g3
library(foreign)
read.dta("~/framshare/bmi.g3.dta")->bmi
subset(bmi,select=c("shareid","height","weight"))->bmi
merge(g3,bmi,by.x="id",by.y="shareid")->g3


#Pairwise means and SDs yes, for height, for bmi, for education, and for age/sex.
#But then we have to do the same pairwise counts of minor alleles and mean principle components for that pair so that we can run the analysis ultimately...

pair.fun<-function(x,fun.nm,var,extra.arg) {
  x[order(x$id),]->x
  tmp<-list()
  nrow(x)->N
  for (i in 1:(N-1)) {
    index2<-(i+1):N
    index1<-rep(i,length(index2))
    data.frame(paste(x$id[index1],x$id[index2],sep="_"),x[[var]][index1],x[[var]][index2])->tmp[[i]]
  }
  do.call("rbind",tmp)->df
  c('pair.id',paste(var,"1",sep="_"),paste(var,"2",sep="_"))->names(df)
  get(fun.nm)->fun
  apply(matrix(unlist(df[,2:3]),ncol=2,byrow=FALSE),1,fun)->df[[paste(fun.nm,var,sep="_")]]
  df
}
split(g3,g3$family_id)->fams
sapply(fams,nrow)->index
fams[index>1]->fams

out<-list()
for (var in c("bmi","height","bmi","education","age","sex")) {
  lapply(fams,pair.fun,fun.nm='mean',var=var)->v.mean
  lapply(fams,pair.fun,fun.nm='sd',var=var)->v.sd
  do.call("rbind",v.mean)->out[[paste(var,'mean',sep="_")]]
  do.call("rbind",v.sd)->v.sd
  v.sd[,c(1,4)]->out[[paste(var,'sd',sep="_")]]
}

list.merge<-function(L,by="pair.id") {
  L[[1]]->out
  for (i in 2:length(L)) merge(out,L[[i]],by=by,all=TRUE)->out
  out
}
list.merge(out)->pairs

for (ii in 1:length(fams)) {
  fams[[ii]]->x
  x[order(x$id),]->x
  tmp<-list()
  for (i in 1:(nrow(x)-1)) {
    data.frame(paste(x$id[i],x$id[-i],sep="_"),names(fams)[ii])->tmp[[i]]
  }
  do.call("rbind",tmp)->tmp
  c('pair.id','family_id')->names(tmp)
  tmp->fams[[ii]]
}
do.call("rbind",fams)->fams
merge(pairs,fams,by="pair.id")->pairs

#save(pairs,file="/tmp/pairs_data.Rdata")
save(pairs,file="/framingham/pairs_data.Rdata")