# weighted kernel density estimation fast weighted kernel density estimation for 2-dimension and calling C function to implement the calculation procedure ```r kdeC(x,H,gridsize,cutNum,w) ``` ## Arguments - `x`: data points in the format n*2 matrix - `H`: bandwidth, a vector containing 2 num values and set c(0.01,0.01) as default - `gridsize`: number of points for each direction, a vector containing 2 int values and set c(200,50) as default - `cutNum`: number of pieces to be cutted for each direction, a vector containing 2 int values and set c(1,1) as default - `w`: weight, a vector corresponding to parameter `x` and set rep(1,length(x)/2) as default ## Details The function `kdeC` is only suitable for 2-dimension data. The advantage of `kdeC` is that it can get the result quickly because the calculation procedure is implemented in C code. ## Returns the returned value is a list - **estimate**: density estimate at points `evalpointsX` and `evalpointsY` - **evalpointsX**: points at which the `estimate` is evaluated at x-axis direction - **evalpointsY**: points at which the `estimate` is evaluated at y-axis direction ## References R package 'ks' ## Author(s) Kunyu Ye ## Examples ```r data.gen<-function(n.peaks=100, N=1e5, max.var=0.001, max.corr=0.5) { library(mvtnorm) dat<-matrix(0, nrow=N, ncol=2) all.m<-c(NA,NA) for(i in 1:n.peaks) { this.m<-runif(2) this.var<-runif(2, min=0.1*max.var, max=max.var) this.cov<-runif(1, min=-1*max.corr, max=max.corr) * sqrt(this.var[1])* sqrt(this.var[2]) this.s<-matrix(c(this.var[1], this.cov, this.cov, this.var[2]),ncol=2) dat[((i-1)*N/n.peaks+1):(i*N/n.peaks),]<-rmvnorm(N/n.peaks, mean=this.m, sigma=this.s) all.m<-rbind(all.m, this.m) } all.m[,1]<-(all.m[,1]-min(dat[,1]))/diff(range(dat[,1])) all.m[,2]<-(all.m[,2]-min(dat[,2]))/diff(range(dat[,2])) dat[,1]<-(dat[,1]-min(dat[,1]))/diff(range(dat[,1])) dat[,2]<-(dat[,2]-min(dat[,2]))/diff(range(dat[,2])) all.m<-all.m[-1,] return(list(dat=dat,m=all.m)) } r<-data.gen(n.peaks=100, N=1e5, max.var=0.001, max.corr=0.5) k1<-kdeC(r$dat, H=c(0.005,0.005), gridsize = c(501,501), cutNum=c(1,1)) k2<-kdeC(r$dat, H=c(0.005,0.005), gridsize = c(101,101), cutNum=c(5,5)) ```

kdeC function