cents (version 0.1-41)

NiagaraToxic: Successive readings of a toxic substance in the Niagara River near Fort Erie, Ontario.

Description

Niagara River at Fort Erie, successive readings of 12-Dichloro in units of ng/L measured approximately biweekly.

Usage

data(NiagaraToxic)

Format

A data frame with 144 observations on the following 4 variables.

toxic

a numeric vector

jday

a numeric vector

cQ

a numeric vector

cL

a numeric vector

Source

Dr. Abdel El-Shaarwai, Environment Canada

Details

Dr. Abdel El-Shaarwai provided through Environment Canada some a special water quality time series that is of great practical interest. The time series is from Station ON02HA0019 (Fort Erie) on the water quality of the Niagara River. There are more than 500 water quality parameters or variables of interest in this river. The water quality in this river is montiored by a joint U.S./Canada committee. One important toxic variable of great interest is a chemical known as 12-Dichloro which when dissolved in water is measured in units of ng/L. We use a portion of the recent data on this variable that was measured approximately every two weeks over the period from March 1, 2001. This period was chosen because it is the most recent period over which we have a time series of approximately biweekly observations. The time series plot below plots the Julian day number defined so that Julian day number 1 corresponds to the date of the first observation (March 1, 2001).

In total there are 144 values and the data are left censored. The observed censoring rate is r=21/144=14 After March 24, 2005 the detection level for 12 Dichloro dropped from 0.214 to 0.0878. After this change there was only one censored value at Julian day number 1807. Before the change in censoring there were 75 complete observations and 20 censored ones while from March 24, 2005 to the last observation on March 22, 2007 there were 48 complete observations and only one censored observation.

References

N. M. Mohammad (2014). Censored time series analysis. Ph.D. Thesis, Western University.

Examples

data(NiagaraToxic)
head(NiagaraToxic)

#Example from thesis
## Not run: 
# #Diagnostic checks and bootstrap confidence intervals
# Zdf <- NiagaraTmeans <- apply(A, MARGIN=2, fun=mean)oxic
# z <- log(Zdf$toxic)
# iz <- c("o", "L")[1+Zdf$cQ]
# #
# #CENARMA(1,1) 
# cenarma(z, iz, p=1, q=1)
# #fit CENAR(1)
# cenarma(z, iz, p=1)
# #
# #diagnostic checks########
# #test CENARMA(1,1)
# SimModel <- function(OUTCENARMA){
#   outSim <- boot.Cenarma(OUTCENARMA)
# }
# FitModel <- function(outSim){
#   z <- outSim$y
#   iz <- outSim$iy
#   ans <- cenarma(z, iz, p=1, q=1)
#   res <- resid(ans$outarima)
#   list(res=res)
# }
# OUTCENARMA <- cenarma(y=NiagaraToxic$toxic, iy=c("o", "L")[1+NiagaraToxic$cQ], p=1, q=1)
# func <- list(SimModel=SimModel, FitModel=FitModel)
# start.time <- proc.time()[3]
# outp <- portest(OUTCENARMA$outarima, lags=5:25, nslaves=8, NREP=10^3, func=func, test="LjungBox")
# total.time <- proc.time()[3]-start.time
# total.time
# plot(outp[,1], outp[,4], xlab="lag", ylab="P-Value", cex=1.5, col="blue", pch=18, ylim=c(0,1))
# abline(col="red", h=0.05)
# #
# #test CENAR(1)
# SimModel <- function(OUTCENARMA){
#   boot.Cenarma(OUTCENARMA)
# }
# FitModel <- function(outSim){
#   z <- outSim$y
#   iz <- outSim$iy
#   ans <- cenarma(z, iz, p=1)
#   res <- resid(ans$outarima)
#   list(res=res)
# }
# OUTCENARMA <- cenarma(y=log(NiagaraToxic$toxic), iy=c("o", "L")[1+NiagaraToxic$cQ], p=1)
# func <- list(SimModel=SimModel, FitModel=FitModel)
# start.time <- proc.time()[3]
# outp <- portest(OUTCENARMA$outarima, lags=5:25, nslaves=8, NREP=10^3, func=func, test="LjungBox")
# total.time <- proc.time()[3]-start.time
# total.time
# plot(outp[,1], outp[,4], xlab="lag", ylab="P-Value", cex=1.5, col="blue", pch=18, ylim=c(0,1))
# abline(col="red", h=0.05)
# #
# #bootstrap confidence intervals
# #CENARMA(1,1)
# OUTCENARMA <- cenarma(y=log(NiagaraToxic$toxic), iy=c("o", "L")[1+NiagaraToxic$cQ], p=1, q=1)
# nboot <- 1000 
# A <- matrix(numeric(nboot*3), ncol=3)
# start.time <- proc.time()[3]
# for (iboot in 1:nboot){
#   out <- boot.Cenarma(OUTCENARMA)
#   A[iboot, ] <- coef(cenarma(y=out$y, iy=out$iy, p=1, q=1)$outarima)
# }
# total.time <- proc.time()[3]-start.time
# total.time
# means <- apply(A, MARGIN=2, FUN=mean)
# means
# LO <- apply(A, MARGIN=2, function(x) quantile(x, 0.025))
# HI <- apply(A, MARGIN=2, function(x) quantile(x, 0.975))
# ansARMA11 <- matrix(c(LO,HI), nrow=3, dimnames=list(c("phi","theta","mu"), c("lo", "hi")))
# #CEAR(1)
# OUTCENARMA <- cenarma(y=log(NiagaraToxic$toxic), iy=c("o", "L")[1+NiagaraToxic$cQ], p=1)
# nboot <- 1000 
# A <- matrix(numeric(nboot*2), ncol=2)
# start.time <- proc.time()[3]
# for (iboot in 1:nboot){
#   out <- boot.Cenarma(OUTCENARMA)
#   A[iboot, ] <- coef(cenarma(y=out$y, iy=out$iy, p=1)$outarima)
# }
# total.time <- proc.time()[3]-start.time
# total.time
# means <- apply(A, MARGIN=2, FUN=mean)
# means
# LO <- apply(A, MARGIN=2, function(x) quantile(x, 0.025))
# HI <- apply(A, MARGIN=2, function(x) quantile(x, 0.975))
# ansAR2 <- matrix(c(LO,HI), nrow=2, dimnames=list(c("phi","mu"), c("lo", "hi")))
# #summary
# ansARMA11
# ansAR2
# ## End(Not run)