R/find_t_cutoff.R

In TimothyHyndman/deconvolve: Deconvolution Tools for Measurement Error Problems

find_t_cutoff <- function(phi_U, t_phi_U,n){
	
	# Find smallest t > 0 at which phi_U reaches its largest local maximum
	ind <- which(t_phi_U >= 0)
	d_phi_U <- phi_U[ind[2:length(ind)]] - phi_U[ind[1:(length(ind) - 1)]]

	if (length(which(d_phi_U >= 0)) == 0) {
		# phi_U is always decreasing for positive t
		t_cutoff <- t_phi_U[length(t_phi_U)]
	} else {
		first_min_ind <- ind[min(which(d_phi_U >= 0))]
		phi_U_threshold <- max(phi_U[ind[ind>=first_min_ind]])
		tmp <- t_phi_U[phi_U <= phi_U_threshold]
		t_cutoff <- min(tmp[tmp > 0])	
	}	
	
	#Index of smallest t>0 where phi_U(t) <n^(-1/2)
	ind2=length(ind)
	testphi=which(phi_U[ind] < n^(-1/2))
	if(length(testphi)>0){
		ind2 <- min(testphi)
	}
	t_cutoff<-min(t_cutoff,t_phi_U[ind[ind2]])
	t_cutoff
}