Nothing
R/cpt.class.R
In changepoint: Methods for Changepoint Detection
setClass("cpt",slots=list(data.set="ts", cpttype="character", method="character", test.stat="character",pen.type="character",pen.value="numeric",minseglen="numeric",cpts="numeric",ncpts.max="numeric",param.est="list",date="character",version="character"),prototype=prototype(cpttype="Not Set",date=date(),version=as(packageVersion("changepoint"),'character')))
setClass("cpt.reg",slots=list(data.set="matrix", cpttype="character", method="character", test.stat="character",pen.type="character",pen.value="numeric",minseglen="numeric",cpts="numeric",ncpts.max="numeric",param.est="list",date="character",version="character"),prototype=prototype(cpttype="regression",date=date(),version=as(packageVersion("changepoint"),"character")))
# setClass("cpt", representation(), prototype())
# # cpts is the optimal segementation
#
setClass("cpt.range",slots=list(cpts.full="matrix", pen.value.full="numeric"), prototype=prototype(), contains="cpt")
# cpts.full is the entire matrix
# pen.value.full (beta) values as an extra slot (vector)
# retrival functions for slots
if(!isGeneric("data.set")) {
if (is.function("data.set")){
fun <- data.set
}
else {fun <- function(object){
standardGeneric("data.set")
}
}
setGeneric("data.set", fun)
}
setMethod("data.set","cpt",function(object) coredata(object@data.set))
setMethod("data.set","cpt.reg",function(object) coredata(object@data.set))
if(!isGeneric("data.set.ts")) {
if (is.function("data.set.ts")){
fun <- data.set.ts
}
else {fun <- function(object){
standardGeneric("data.set.ts")
}
}
setGeneric("data.set.ts", fun)
}
setMethod("data.set.ts","cpt",function(object) object@data.set)
if(!isGeneric("cpttype")) {
if (is.function("cpttype")){
fun <- cpttype
}
else {fun <- function(object){
standardGeneric("cpttype")
}
}
setGeneric("cpttype", fun)
}
setMethod("cpttype","cpt",function(object) object@cpttype)
setMethod("cpttype","cpt.reg",function(object) object@cpttype)
if(!isGeneric("method")) {
if (is.function("method")){
fun <- method
}
else {fun <- function(object){
standardGeneric("method")
}
}
setGeneric("method", fun)
}
setMethod("method","cpt",function(object) object@method)
setMethod("method","cpt.reg",function(object) object@method)
# distribution remains for backwards compatability, changed to test.stat version 1.0
if(!isGeneric("distribution")) {
if (is.function("distribution")){
fun <- distribution
}
else {fun <- function(object){
standardGeneric("distribution")
}
}
setGeneric("distribution", fun)
}
setMethod("distribution","cpt",function(object) object@test.stat)
setMethod("distribution","cpt.reg",function(object) object@test.stat)
if(!isGeneric("test.stat")) {
if (is.function("test.stat")){
fun <- test.stat
}
else {fun <- function(object){
standardGeneric("test.stat")
}
}
setGeneric("test.stat", fun)
}
setMethod("test.stat","cpt",function(object) object@test.stat)
setMethod("test.stat","cpt.reg",function(object) object@test.stat)
if(!isGeneric("pen.type")) {
if (is.function("pen.type")){
fun <- pen.type
}
else {fun <- function(object){
standardGeneric("pen.type")
}
}
setGeneric("pen.type", fun)
}
setMethod("pen.type","cpt",function(object) object@pen.type)
setMethod("pen.type","cpt.reg",function(object) object@pen.type)
if(!isGeneric("pen.value")) {
if (is.function("pen.value")){
fun <- pen.value
}
else {fun <- function(object){
standardGeneric("pen.value")
}
}
setGeneric("pen.value", fun)
}
setMethod("pen.value","cpt",function(object) object@pen.value)
setMethod("pen.value","cpt.reg",function(object) object@pen.value)
if(!isGeneric("pen.value.full")) {
if (is.function("pen.value.full")){
fun <- pen.value.full
}
else {fun <- function(object){
standardGeneric("pen.value.full")
}
}
setGeneric("pen.value.full", fun)
}
setMethod("pen.value.full","cpt.range",function(object) object@pen.value.full)
if(!isGeneric("minseglen")) {
if (is.function("minseglen")){
fun <- minseglen
}
else {fun <- function(object){
standardGeneric("minseglen")
}
}
setGeneric("minseglen", fun)
}
setMethod("minseglen","cpt",function(object) object@minseglen)
if(!isGeneric("cpts")) {
if (is.function("cpts")){
fun <- cpts
}
else {fun <- function(object,...){
standardGeneric("cpts")
}
}
setGeneric("cpts", fun)
}
setMethod("cpts","cpt",function(object) object@cpts[-length(object@cpts)])
setMethod("cpts","cpt.reg",function(object) object@cpts[-length(object@cpts)])
setMethod("cpts","cpt.range",function(object,ncpts=NA){
if(is.na(ncpts)){return(object@cpts[-length(object@cpts)])}
else{
ncpts.full=apply(cpts.full(object),1,function(x){sum(x>0,na.rm=TRUE)})
row=try(which(ncpts.full==ncpts),silent=TRUE)
if(inherits(row,'try-error')){
stop("Your input object doesn't have a segmentation with the requested number of changepoints.\n Possible ncpts are: ",paste(ncpts.full,collapse=','))
}
else{return(cpts.full(object)[row,])}
}
})
if(!isGeneric("cpts.full")) {
if (is.function("cpts.full")){
fun <- cpts.full
}
else {fun <- function(object){
standardGeneric("cpts.full")
}
}
setGeneric("cpts.full", fun)
}
setMethod("cpts.full","cpt.range",function(object) object@cpts.full)
if(!isGeneric("cpts.ts")) {
if (is.function("cpts.ts")){
fun <- cpts.ts
}
else {fun <- function(object){
standardGeneric("cpts.ts")
}
}
setGeneric("cpts.ts", fun)
}
setMethod("cpts.ts","cpt",function(object) index(data.set.ts(object))[cpts(object)] )
if(!isGeneric("ncpts.max")) {
if (is.function("ncpts.max")){
fun <- ncpts.max
}
else {fun <- function(object){
standardGeneric("ncpts.max")
}
}
setGeneric("ncpts.max", fun)
}
setMethod("ncpts.max","cpt",function(object) object@ncpts.max)
setMethod("ncpts.max","cpt.reg",function(object) object@ncpts.max)
if(!isGeneric("param.est")) {
if (is.function("param.est")){
fun <- param.est
}
else {fun <- function(object){
standardGeneric("param.est")
}
}
setGeneric("param.est", fun)
}
setMethod("param.est","cpt",function(object) object@param.est)
setMethod("param.est","cpt.reg",function(object) object@param.est)
setMethod("coef","cpt",function(object) object@param.est)
setMethod("coef","cpt.reg",function(object) object@param.est)
# ncpts function
if(!isGeneric("ncpts")) {
if (is.function("ncpts")){
fun <- ncpts
}
else {fun <- function(object){
standardGeneric("ncpts")
}
}
setGeneric("ncpts", fun)
}
setMethod("ncpts","cpt",function(object) length(cpts(object)))
setMethod("ncpts","cpt.reg",function(object) length(cpts(object)))
# seg.len function
if(!isGeneric("seg.len")) {
if (is.function("seg.len")){
fun <- seg.len
}
else {fun <- function(object){
standardGeneric("seg.len")
}
}
setGeneric("seg.len", fun)
}
setMethod("seg.len","cpt",function(object){object@cpts-c(0,object@cpts[-length(object@cpts)])})
setMethod("seg.len","cpt.reg",function(object){object@cpts-c(0,object@cpts[-length(object@cpts)])})
#i.e. if there is a changepoint in the data, return segment length. If not, return length of the data
# nseg function
if(!isGeneric("nseg")) {
if (is.function("nseg")){
fun <- nseg
}
else {fun <- function(object){
standardGeneric("nseg")
}
}
setGeneric("nseg", fun)
}
setMethod("nseg","cpt",function(object){ncpts(object)+1})
setMethod("nseg","cpt.reg",function(object){ncpts(object)+1})
# replacement functions for slots
setGeneric("data.set<-", function(object, value) standardGeneric("data.set<-"))
setReplaceMethod("data.set", "cpt", function(object, value) {
if(is.ts(value)){object@data.set <- value}else{object@data.set <- ts(value)}
return(object)
})
setReplaceMethod("data.set", "cpt.reg", function(object, value) {
object@data.set <- value
return(object)
})
setGeneric("cpttype<-", function(object, value) standardGeneric("cpttype<-"))
setReplaceMethod("cpttype", "cpt", function(object, value) {
object@cpttype <- value
return(object)
})
setReplaceMethod("cpttype", "cpt.reg", function(object, value) {
object@cpttype <- value
return(object)
})
setGeneric("method<-", function(object, value) standardGeneric("method<-"))
setReplaceMethod("method", "cpt", function(object, value) {
object@method <- value
return(object)
})
setReplaceMethod("method", "cpt.reg", function(object, value) {
object@method <- value
return(object)
})
# distribution remains for backwards compatability, changed to test.stat version 1.0
setGeneric("distribution<-", function(object, value) standardGeneric("distribution<-"))
setReplaceMethod("distribution", "cpt", function(object, value) {
object@test.stat <- value
return(object)
})
setReplaceMethod("distribution", "cpt.reg", function(object, value) {
object@test.stat <- value
return(object)
})
setGeneric("test.stat<-", function(object, value) standardGeneric("test.stat<-"))
setReplaceMethod("test.stat", "cpt", function(object, value) {
object@test.stat <- value
return(object)
})
setReplaceMethod("test.stat", "cpt.reg", function(object, value) {
object@test.stat <- value
return(object)
})
setGeneric("pen.type<-", function(object, value) standardGeneric("pen.type<-"))
setReplaceMethod("pen.type", "cpt", function(object, value) {
object@pen.type <- value
return(object)
})
setReplaceMethod("pen.type", "cpt.reg", function(object, value) {
object@pen.type <- value
return(object)
})
setGeneric("pen.value<-", function(object, value) standardGeneric("pen.value<-"))
setReplaceMethod("pen.value", "cpt", function(object, value) {
object@pen.value <- value
return(object)
})
setReplaceMethod("pen.value", "cpt.reg", function(object, value) {
object@pen.value <- value
return(object)
})
setGeneric("minseglen<-", function(object, value) standardGeneric("minseglen<-"))
setReplaceMethod("minseglen", "cpt", function(object, value) {
object@minseglen <- value
return(object)
})
setReplaceMethod("minseglen", "cpt.range", function(object, value) {
object@minseglen <- value
return(object)
})
setReplaceMethod("minseglen", "cpt.reg", function(object, value) {
object@minseglen <- value
return(object)
})
setGeneric("cpts<-", function(object, value) standardGeneric("cpts<-"))
setReplaceMethod("cpts", "cpt", function(object, value) {
if((cpttype(object)=="meanar")|(cpttype(object)=="trendar")){
n=length(object@data.set)-1
}
else{n=length(object@data.set)}
if(value[length(value)]==n){object@cpts <- value}
else{ object@cpts <- c(value,n) }
return(object)
})
setReplaceMethod("cpts", "cpt.range", function(object, value) {
if((cpttype(object)=="meanar")|(cpttype(object)=="trendar")){
n=length(object@data.set)-1
}
else{n=length(object@data.set)}
if(value[length(value)]==n){object@cpts <- value}
else{ object@cpts <- c(value,n) }
return(object)
})
setReplaceMethod("cpts", "cpt.reg", function(object, value) {
if(value[length(value)]==nrow(object@data.set)){object@cpts <- value}
else{ object@cpts <- c(value,nrow(object@data.set)) }
return(object)
})
setGeneric("ncpts.max<-", function(object, value) standardGeneric("ncpts.max<-"))
setReplaceMethod("ncpts.max", "cpt", function(object, value) {
object@ncpts.max <- value
return(object)
})
setReplaceMethod("ncpts.max", "cpt.reg", function(object, value) {
object@ncpts.max <- value
return(object)
})
setGeneric("param.est<-", function(object, value) standardGeneric("param.est<-"))
setReplaceMethod("param.est", "cpt", function(object, value) {
object@param.est <- value
return(object)
})
setReplaceMethod("param.est", "cpt.reg", function(object, value) {
object@param.est <- value
return(object)
})
setGeneric("cpts.full<-", function(object, value) standardGeneric("cpts.full<-"))
setReplaceMethod("cpts.full", "cpt.range", function(object, value) {
object@cpts.full <- value
return(object)
})
setGeneric("pen.value.full<-", function(object, value) standardGeneric("pen.value.full<-"))
setReplaceMethod("pen.value.full", "cpt.range", function(object, value) {
object@pen.value.full <- value
return(object)
})
# setGeneric("pen.value.input<-", function(object, value) standardGeneric("pen.value.input<-"))
# setReplaceMethod("pen.value.input", "cpt", function(object, value) {
# object@pen.value.input <- value
# return(object)
# })
# parameter functions
setGeneric("param", function(object,...) standardGeneric("param"))
setMethod("param", "cpt", function(object,shape,...) {
param.mean=function(object){
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
data=data.set(object)
tmpmean=NULL
for(j in 1:nseg(object)){
tmpmean[j]=mean(data[(cpts[j]+1):(cpts[j+1])])
}
return(tmpmean)
}
param.var=function(object){
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
data=data.set(object)
seglen=seg.len(object)
tmpvar=NULL
for(j in 1:nseg(object)){
tmpvar[j]=var(data[(cpts[j]+1):(cpts[j+1])])
}
tmpvar=tmpvar*(seglen-1)/seglen # correctly for the fact that the MLE estimate is /n but the var function is /n-1
return(tmpvar)
}
param.scale=function(object,shape){
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
data=data.set(object)
y=c(0,cumsum(data))
tmpscale=NULL
for(j in 1:nseg(object)){
tmpscale[j]=(y[(cpts[j+1]+1)]-y[(cpts[j]+1)])/((cpts[j+1]-cpts[j])*shape)
}
return(tmpscale)
}
param.trend=function(object){
cpts=c(0,object@cpts)
seglen=seg.len(object)
data=data.set(object)
n=length(data)
sumstat=cbind(cumsum(c(0,data)),cumsum(c(0,data*c(1:n))))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=2)
cptsumstat[,2]=cptsumstat[,2]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx3
thetaS=(2*cptsumstat[,1]*(2*seglen + 1) - 6*cptsumstat[,2]) / (2*seglen*(2*seglen + 1) - 3*seglen*(seglen+1))
thetaT=(6*cptsumstat[,2])/((seglen+1)*(2*seglen+1)) + (thetaS * (1-((3*seglen)/((2*seglen)+1))))
return(cbind(thetaS,thetaT))
}
param.meanar=function(object){
seglen=seg.len(object)
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]^2)),cumsum(c(0,data[-(n+1)]^2)))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=5)
beta2=(2*seglen*cptsumstat[,3]-cptsumstat[,1]*cptsumstat[,2])/(2*seglen*cptsumstat[,5]*(1-cptsumstat[,2]^2));
beta1=(2*cptsumstat[,1]-beta2*cptsumstat[,2])/(2*seglen);
return(cbind(beta1,beta2))
}
param.trendar=function(object){
seglen=seg.len(object)
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]*c(1:n))),cumsum(c(0,data[-(n+1)]*c(0:(n-1)))),cumsum(c(0,data[-1]^2)),cumsum(c(0,data[-(n+1)]^2)))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=7)
cptsumstat[,4]=cptsumstat[,4]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx4
cptsumstat[,5]=cptsumstat[,5]-cptsumstat[,2]*c(0,cpts(object)) # i.e. creating newx5
betatop=seglen*(seglen-1)*(seglen*(seglen-1)*cptsumstat[,3] + 2*(2*seglen+1)*cptsumstat[,1]*(cptsumstat[,5]-seglen*cptsumstat[,2]) + 6*cptsumstat[,4]*(cptsumstat[,2]-cptsumstat[,5]))
betabottom=seglen*(seglen-1)*cptsumstat[,7] + 2*(2*seglen+1)*cptsumstat[,2]*(seglen*cptsumstat[,2]-cptsumstat[,5]) + 6*cptsumstat[,5]*(cptsumstat[,5]-cptsumstat[,2]);
beta=betatop/betabottom;
thetajpo=(6*(seglen+2)*(cptsumstat[,4]-beta*cptsumstat[,5]))/((seglen+1)*(2*seglen+1)) - 2*(cptsumstat[,1]-beta*cptsumstat[,2])
thetaj=(2*(2*seglen+1)*(cptsumstat[,1]-beta*cptsumstat[,2])-6*(cptsumstat[,4]-beta*cptsumstat[,5]))/(seglen-1)
return(cbind(beta,thetajpo,thetaj))
}
if(cpttype(object)=="mean"){
param.est(object)<-list(mean=param.mean(object))
}
else if(cpttype(object)=="variance"){
param.est(object)<-list(variance=param.var(object))
}
else if(cpttype(object)=="mean and variance"){
if(test.stat(object)=="Normal"){
param.est(object)<-list(mean=param.mean(object),variance=param.var(object))
}
else if(test.stat(object)=="Gamma"){
param.est(object)<-list(scale=param.scale(object,shape=shape),shape=shape)
}
else if(test.stat(object)=="Exponential"){
param.est(object)<-list(rate=1/param.mean(object))
}
else if(test.stat(object)=="Poisson"){
param.est(object)<-list(lambda=param.mean(object))
}
else{
stop("Unknown test statistic for a change in mean and variance")
}
}
else if(cpttype(object)=="trend"){
if(test.stat(object)=="Normal"){
tmp=param.trend(object)
param.est(object)<-list(thetaS=tmp[,1],thetaT=tmp[,2])
}
else{
stop("Unknown test statistic for a change in trend")
}
}
else if(cpttype(object)=="trendar"){
if(test.stat(object)=="Normal"){
tmp=param.trendar(object)
param.est(object)<-list(beta=tmp[,1],thetajpo=tmp[,2],thetaj=tmp[,3])
}
else{
stop("Unknown test statistic for a change in trend+ar")
}
}
else if(cpttype(object)=="meanar"){
if(test.stat(object)=="Normal"){
tmp=param.meanar(object)
param.est(object)<-list(beta1=tmp[,1],beta2=tmp[,2])
}
else{
stop("Unknown test statistic for a change in mean+ar")
}
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance', 'mean and variance', 'trend', 'meanar' or 'trendar'.")
}
return(object)
})
setMethod("param", "cpt.range", function(object,ncpts=NA,shape,...) {
if(is.na(ncpts)){
cpts=object@cpts
if(cpts[1]!=0){cpts=c(0,cpts)} # PELT derivatives don't include the 0
}
else{
ncpts.full=apply(cpts.full(object),1,function(x){sum(x>0,na.rm=TRUE)})
row=try(which(ncpts.full==ncpts),silent=TRUE)
if(inherits(row,'try-error')){
stop("Your input object doesn't have a segmentation with the requested number of changepoints.")
}
cpts=c(0,cpts.full(object)[row,1:ncpts],length(data.set(object)))
}
param.mean=function(object,cpts){
nseg=length(cpts)-1
data=data.set(object)
tmpmean=NULL
for(j in 1:nseg){
tmpmean[j]=mean(data[(cpts[j]+1):(cpts[j+1])])
}
return(tmpmean)
}
param.var=function(object,cpts){
nseg=length(cpts)-1
data=data.set(object)
seglen=cpts[-1]-cpts[-length(cpts)]
tmpvar=NULL
for(j in 1:nseg){
tmpvar[j]=var(data[(cpts[j]+1):(cpts[j+1])])
}
tmpvar=tmpvar*(seglen-1)/seglen
return(tmpvar)
}
param.scale=function(object,cpts,shape){
nseg=length(cpts)-1
data=data.set(object)
y=c(0,cumsum(data))
tmpscale=NULL
for(j in 1:nseg){
tmpscale[j]=(y[(cpts[j+1]+1)]-y[(cpts[j]+1)])/((cpts[j+1]-cpts[j])*shape)
}
return(tmpscale)
}
param.trend=function(object){
cpts=c(0,object@cpts)
seglen=seg.len(object)
data=data.set(object)
n=length(data)
sumstat=cbind(cumsum(c(0,data)),cumsum(c(0,data*c(1:n))))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=2)
cptsumstat[,2]=cptsumstat[,2]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx3
thetaS=(2*cptsumstat[,1]*(2*seglen + 1) - 6*cptsumstat[,2]) / (2*seglen*(2*seglen + 1) - 3*seglen*(seglen+1))
thetaT=(6*cptsumstat[,2])/((seglen+1)*(2*seglen+1)) + (thetaS * (1-((3*seglen)/((2*seglen)+1))))
return(cbind(thetaS,thetaT))
}
param.meanar=function(object,cpts){
seglen=cpts[-1]-cpts[-length(cpts)]
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]^2)),cumsum(c(0,data[-(n+1)]^2)))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=5)
beta2=(2*seglen*cptsumstat[,3]-cptsumstat[,1]*cptsumstat[,2])/(2*seglen*cptsumstat[,5]*(1-cptsumstat[,2]^2));
beta1=(2*cptsumstat[,1]-beta2*cptsumstat[,2])/(2*seglen);
return(cbind(beta1,beta2))
}
param.trendar=function(object,cpts){
seglen=cpts[-1]-cpts[-length(cpts)]
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]*c(1:n))),cumsum(c(0,data[-(n+1)]*c(0:(n-1)))))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=7)
cptsumstat[,4]=cptsumstat[,4]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx4
cptsumstat[,5]=cptsumstat[,5]-cptsumstat[,2]*c(0,cpts(object)) # i.e. creating newx5
betatop=seglen*(seglen-1)*(seglen*(seglen-1)*cptsumstat[,3] + 2*(2*seglen+1)*cptsumstat[,1]*(cptsumstat[,5]-seglen*cptsumstat[,2]) + 6*cptsumstat[,4]*(cptsumstat[,2]-cptsumstat[,5]))
betabottom=seglen*(seglen-1)*cptsumstat[,7] + 2*(2*seglen+1)*cptsumstat[,2]*(seglen*cptsumstat[,2]-cptsumstat[,5]) + 6*cptsumstat[,5]*(cptsumstat[,5]-cptsumstat[,2]);
beta=betatop/betabottom;
thetajpo=(6*(seglen+2)*(cptsumstat[,4]-beta*cptsumstat[,5]))/((seglen+1)*(2*seglen+1)) - 2*(cptsumstat[,1]-beta*cptsumstat[,2])
thetaj=(2*(2*seglen+1)*(cptsumstat[,1]-beta*cptsumstat[,2])-6*(cptsumstat[,4]-beta*cptsumstat[,5]))/(seglen-1)
return(cbind(beta,thetajpo,thetaj))
}
if(cpttype(object)=="mean"){
param.est<-list(mean=param.mean(object,cpts))
}
else if(cpttype(object)=="variance"){
param.est<-list(variance=param.var(object,cpts))
}
else if(cpttype(object)=="mean and variance"){
if(test.stat(object)=="Normal"){
param.est<-list(mean=param.mean(object,cpts),variance=param.var(object,cpts))
}
else if(test.stat(object)=="Gamma"){
param.est<-list(scale=param.scale(object,cpts,shape=shape),shape=shape)
}
else if(test.stat(object)=="Exponential"){
param.est<-list(rate=1/param.mean(object,cpts))
}
else if(test.stat(object)=="Poisson"){
param.est<-list(lambda=param.mean(object,cpts))
}
else{
stop("Unknown test statistic for a change in mean and variance")
}
}
else if(cpttype(object)=="trend"){
if(test.stat(object)=="Normal"){
tmp=param.trend(object)
param.est(object)<-list(thetaS=tmp[,1],thetaT=tmp[,2])
}
else{
stop("Unknown test statistic for a change in trend")
}
}
else if(cpttype(object)=="trendar"){
if(test.stat(object)=="Normal"){
tmp=param.trendar(object)
param.est(object)<-list(beta=tmp[,1],thetajpo=tmp[,2],thetaj=tmp[,3])
}
else{
stop("Unknown test statistic for a change in trend+ar")
}
}
else if(cpttype(object)=="meanar"){
if(test.stat(object)=="Normal"){
tmp=param.meanar(object)
param.est(object)<-list(beta1=tmp[,1],beta2=tmp[,2])
}
else{
stop("Unknown test statistic for a change in mean+ar")
}
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance', 'mean and variance', 'trend', 'meanar' or 'trendar'")
}
if(is.na(ncpts)){
param.est(object)=param.est
return(object)
}
out=new('cpt.range')
param.est(out)=param.est
return(out)
})
setMethod("param", "cpt.reg", function(object,shape,...) {
param.norm=function(object){
cpts=c(0,object@cpts)
# nseg=length(cpts)-1 #nseg(object)
data=data.set(object)
p=ncol(data)-1
tmpbeta=matrix(NA,ncol=p,nrow=nseg(object))
tmpsigma=rep(NA,nseg(object))
for(j in 1:nseg(object)){
formula=paste('-1+data[',cpts[j]+1,':',cpts[j+1],',2]',sep='')
if(p>1){
for(i in 2:p){
formula=paste(formula,'+data[',(cpts[j]+1),':',cpts[j+1],',',i+1,']',sep='')
}
}
tmpfit=eval(parse(text=paste('lm(data[',(cpts[j]+1),':',cpts[j+1],',1]~',formula,')',sep='')))
tmpbeta[j,]=tmpfit$coefficients
tmpsigma[j]=sum(tmpfit$residuals^2)/(length(tmpfit$residuals)-length(tmpfit$coefficients)) ##var(tmpfit$residuals)
}
return(list(beta=tmpbeta,sig2=tmpsigma))
}
if(test.stat(object)=="Normal"){
param.est(object)<-param.norm(object)
}
else{
stop("Unknown test statistic, must be 'Normal'")
}
return(object)
})
# summary functions
setMethod("summary","cpt",function(object){
cat("Created Using changepoint version",object@version,'\n')
cat("Changepoint type : Change in",cpttype(object),'\n')
cat("Method of analysis :",method(object),"\n")
cat("Test Statistic :", test.stat(object),"\n")
cat("Type of penalty :", pen.type(object), "with value,",pen.value(object),"\n")
cat("Minimum Segment Length :", minseglen(object),"\n")
cat("Maximum no. of cpts :", ncpts.max(object),"\n")
if(length(cpts(object))<=20){cat("Changepoint Locations :",cpts(object),"\n")}
else{cat("Number of changepoints:", ncpts(object),"\n")}
})
setMethod("summary","cpt.range",function(object){
cat("Created Using changepoint version",object@version,'\n')
cat("Changepoint type : Change in",cpttype(object),'\n')
cat("Method of analysis :",method(object),"\n")
cat("Test Statistic :", test.stat(object),"\n")
cat("Type of penalty :", pen.type(object), "with value,",pen.value(object),"\n")
cat("Minimum Segment Length :", minseglen(object),"\n")
cat("Maximum no. of cpts :", ncpts.max(object),"\n")
if(length(cpts(object))<=20){cat("Changepoint Locations :",cpts(object),"\n")}
else{cat("Number of changepoints:", ncpts(object),"\n")}
if((nrow(cpts.full(object))<=5)&(ncol(cpts.full(object)<=20))){cat("Range of segmentations:\n");print(cpts.full(object));cat("\n For penalty values:", pen.value.full(object),"\n")}
else{cat("Number of segmentations recorded:", nrow(cpts.full(object)), " with between ", sum(cpts.full(object)[nrow(cpts.full(object)),]>0,na.rm=T), " and ", sum(cpts.full(object)[1,]>0,na.rm=T), "changepoints.\n Penalty value ranges from:",min(pen.value.full(object))," to ",max(pen.value.full(object)))}
})
setMethod("summary","cpt.reg",function(object){
cat("Created Using changepoint version",object@version,'\n')
cat("Changepoint type : Change in",cpttype(object),'\n')
cat("Method of analysis :",method(object),"\n")
cat("Test Statistic :", test.stat(object),"\n")
cat("Type of penalty :", pen.type(object), "with value,",pen.value(object),"\n")
cat("Maximum no. of cpts :", ncpts.max(object),"\n")
if(length(cpts(object))<=20){cat("Changepoint Locations :",cpts(object),"\n")}
else{cat("Number of changepoints:", ncpts(object),"\n")}
})
# show functions
setMethod("show","cpt",function(object){
cat("Class 'cpt' : Changepoint Object\n")
cat(" ~~ : S4 class containing", length(attributes(object))-1, "slots with names\n")
cat(" ", names(attributes(object))[1:(length(attributes(object))-1)], "\n\n")
cat("Created on :", object@date, "\n\n")
cat("summary(.) :\n----------\n")
summary(object)
})
setMethod("show","cpt.reg",function(object){
cat("Class 'cpt.reg' : Changepoint Regression Object\n")
cat(" ~~ : S4 class containing", length(attributes(object))-1, "slots with names\n")
cat(" ", names(attributes(object))[1:(length(attributes(object))-1)], "\n\n")
cat("Created on :", object@date, "\n\n")
cat("summary(.) :\n----------\n")
summary(object)
})
# plot functions
setMethod("plot","cpt",function(x,cpt.col='red',cpt.width=1,cpt.style=1,...){
if(length(param.est(x))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(x)
cat('done.\n')
}
plot(data.set.ts(x),...)
if(cpttype(x)=="variance" || cpttype(x)=="nonparametric (empirical_distribution)"){
abline(v=index(data.set.ts(x))[cpts(x)],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
else if(cpttype(x)=="mean" || cpttype(x)=="mean and variance"){
#nseg=length(cpts(x))+1
cpts=c(0,x@cpts)
if((test.stat(x)=="Normal")||(test.stat(x)=="CUSUM")){
means=param.est(x)$mean
}
else if(test.stat(x)=="Gamma"){
means=param.est(x)$scale*param.est(x)$shape
}
else if(test.stat(x)=="Exponential"){
means=1/param.est(x)$rate
}
else if(test.stat(x)=="Poisson"){
means=param.est(x)$lambda
}
else{
stop('Invalid Changepoint test statistic')
}
for(i in 1:nseg(x)){
segments(index(data.set.ts(x))[cpts[i]+1],means[i],index(data.set.ts(x))[cpts[i+1]],means[i],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else if(cpttype(x)=="trend"){
cpts=c(0,x@cpts)
intercept=rep(param.est(x)$thetaS,x@cpts-c(0,cpts(x)))
slope=rep(param.est(x)$thetaT-param.est(x)$thetaS,x@cpts-c(0,cpts(x)))/rep(x@cpts-c(0,cpts(x)),x@cpts-c(0,cpts(x)))
cptn=rep(c(0,cpts(x)),x@cpts-c(0,cpts(x)))
n=length(data.set(x))
means=intercept+slope*((1:n)-cptn)
for(i in 1:nseg(x)){
segments(index(data.set.ts(x))[cpts[i]+1],means[cpts[i]+1],index(data.set.ts(x))[cpts[i+1]],means[cpts[i+1]],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else{
stop('Invalid Changepoint Type for plotting.\n Can only plot mean, variance, mean and variance')
}
})
setMethod("plot","cpt.range",function(x,ncpts=NA,diagnostic=FALSE,cpt.col='red',cpt.width=1,cpt.style=1,...){
if(diagnostic==TRUE){
n.changepoints = apply(cpts.full(x), 1, function(x) sum(x > 0, na.rm = TRUE))
penalty.values = pen.value.full(x)
if (is.null(list(...)$type)) {
# By default, the type of the diagnostic plots is "lines".
plot(x = n.changepoints, y = penalty.values, xlab = 'Number of Changepoints', ylab = 'Penalty Value', type = "l", ...)
} else {
plot(x = n.changepoints, y = penalty.values, xlab = 'Number of Changepoints', ylab = 'Penalty Value', ...)
}
return(invisible(NULL))
}
plot(data.set.ts(x),...)
if(is.na(ncpts)){
if(pen.type(x)=="CROPS"){
stop('CROPS does not supply an optimal set of changepoints, set ncpts to the desired segmentation to plot or use diagnostic=TRUE to identify an appropriate number of changepoints')
}
cpts.to.plot=cpts(x)
param.est=x
}
else{
ncpts.full=apply(cpts.full(x),1,function(x){sum(x>0,na.rm=TRUE)})
row=which(ncpts.full==ncpts)
if(length(row)==0){
stop(paste("Your input object doesn't have a segmentation with the requested number of changepoints.\n Possible ncpts are: "),paste(ncpts.full,collapse=','))
}
cpts.to.plot=cpts.full(x)[row,1:ncpts]
if(test.stat(x)=="Gamma"){
param.est=param(x,ncpts,shape=param.est(x)$shape)
}
else{
param.est=param(x,ncpts)
}
}
if(cpttype(x)=="variance"){
abline(v=index(data.set.ts(x))[cpts.to.plot],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
else if(cpttype(x)=="mean" || cpttype(x)=="mean and variance"){
if((test.stat(x)=="Normal")||(test.stat(x)=="CUSUM")){
means=param.est(param.est)$mean
}
else if(test.stat(x)=="Gamma"){
means=param.est(param.est)$scale*param.est(param.est)$shape
}
else if(test.stat(x)=="Exponential"){
means=1/param.est(param.est)$rate
}
else if(test.stat(x)=="Poisson"){
means=param.est(param.est)$lambda
}
else{
stop('Invalid Changepoint test statistic')
}
nseg=length(means)
cpts.to.plot=c(0,cpts.to.plot,length(data.set(x)))
for(i in 1:nseg){
segments(index(data.set.ts(x))[cpts.to.plot[i]+1],means[i],index(data.set.ts(x))[cpts.to.plot[i+1]],means[i],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else{
stop('Invalid Changepoint Type for plotting.\n Can only plot mean, variance, mean and variance')
}
})
setMethod("plot","cpt.reg",function(x,cpt.col='red',cpt.width=1,cpt.style=1,...){
if(length(param.est(x))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(x)
cat('done.\n')
}
plot(data.set(x)[,1],type='l',...)
if(test.stat(x)=="Normal"){
cpts=c(0,x@cpts)
betas=param.est(x)$beta
for(i in 1:nseg(x)){
lines((cpts[i]+1):cpts[i+1],betas[i,]%*%t(data.set(x)[(cpts[i]+1):cpts[i+1],-1]),col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else{
stop('Invalid Changepoint test statistic')
}
})
# likelihood functions
setMethod("logLik", "cpt", function(object) {
if(length(param.est(object))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(object)
cat('done.\n')
}
if(test.stat(object)=="Normal"){
if(cpttype(object)=="mean"){
means=rep(param.est(object)$mean,object@cpts-c(0,cpts(object)))
rss=sum((data.set(object)-means)^2)
n=length(data.set(object))
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
cpts=c(0,object@cpts)
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(cpts[-1]-cpts[-(nseg(object)+1)])))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="variance"){
rss=c(0,cumsum((data.set(object)-param.est(object)$mean)^2))
cpts=c(0,object@cpts)
n=length(data.set(object))
seglen=seg.len(object)
sigmas=(rss[cpts[-1]+1]-rss[cpts[-length(cpts)]+1])/seglen
like=n*log(2*pi)+sum(seglen*log(sigmas))+n
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seglen)))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="mean and variance"){
means=rep(param.est(object)$mean,object@cpts-c(0,cpts(object)))
rss=sum((data.set(object)-means)^2)
n=length(data.set(object))
cpts=c(0,object@cpts)
seglen=seg.len(object)
sigmas=param.est(object)$variance
like=n*log(2*pi)+sum(seglen*log(sigmas))+n
if(pen.type(object)=="MBIC"){
like=c(like,like+(nseg(object)-2)*pen.value(object)+sum(log(seglen)))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="trend"){
intercept=rep(param.est(object)$thetaS,object@cpts-c(0,cpts(object)))
slope=rep(param.est(object)$thetaT-param.est(object)$thetaS,object@cpts-c(0,cpts(object)))/rep(object@cpts-c(0,cpts(object)),object@cpts-c(0,cpts(object)))
cptn=rep(c(0,cpts(object)),object@cpts-c(0,cpts(object)))
n=length(data.set(object))
means=intercept+slope*((1:n)-cptn)
rss=sum((data.set(object)-means)^2)
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="trendar"){
seglen=seg.len(object)
intercept=rep(param.est(object)$thetaj,seglen)
slope=rep(param.est(object)$thetajpo-param.est(object)$thetaj,seglen)/rep(seglen,seglen)
ar=rep(param.est(object)$beta,seglen)
cptn=rep(c(0,cpts(object)),seglen)
n=length(data.set(object))
means=NULL;means[1]=0
for(i in 2:n){means[i]=intercept+slope*((1:n)-cptn)+ar*means[i-1]}
means=means[-1]
rss=sum((data.set(object)[-1]-means)^2)
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="meanar"){
seglen=seg.len(object)
intercept=rep(param.est(object)$beta1,seglen)
ar=rep(param.est(object)$beta2,seglen)
cptn=rep(c(0,cpts(object)),seglen)
n=length(data.set(object))
means[1]=0;for(i in 2:n){means[i]=intercept+ar*means[i-1]}
means=means[-1]
rss=sum((data.set(object)[-1]-means)^2)
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance', 'mean and variance', 'trend', 'meanar' or 'trendar'")
}
}
else if(test.stat(object)=="Gamma"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Gamma', must be 'mean and variance'")
}
else{
warning("Not changed to be -2*logLik")
mll.meanvarg=function(x,n,shape){
return(n*shape*log(n*shape)-n*shape*log(x))
}
y=c(0,cumsum(data.set(object)))
shape=param.est(object)$shape
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg(object)){
tmplike=tmplike+mll.meanvarg(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j],shape)
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg(object)-1)*pen.value(object))
}
}
}
else if(test.stat(object)=="Exponential"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Exponential', must be 'mean and variance'")
}
else{
warning("Not changed to be -2*logLik")
mll.meanvare=function(x,n){
return(n*log(n)-n*log(x))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg(object)){
tmplike=tmplike+mll.meanvare(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg(object)-1)*pen.value(object))
}
}
}
else if(test.stat(object)=="Poisson"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Poisson', must be 'mean and variance'")
}
else{
warning("Not changed to be -2*logLik")
mll.meanvarp=function(x,n){
return(x*log(x)-x*log(n))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg(object)){
tmplike=tmplike+mll.meanvarp(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg(object)-1)*pen.value(object))
}
}
}
else{stop("logLik is only valid for distributional assumptions, not CUSUM or CSS")}
names(like)=c("-2*logLik","-2*Loglike+pen")
return(like)
})
setMethod("logLik", "cpt.range", function(object,ncpts=NA) {
warning("Not changed to be -2*logLik")
if(is.na(ncpts)){
if(pen.type(object)=="CROPS"){
stop('CROPS does not supply an optimal set of changepoints, set ncpts argument to the desired segmentation to plot or use diagnostic=TRUE to identify an appropriate number of changepoints')
}
cpts=c(0,object@cpts)
pen.value=pen.value(object)
}
else{
ncpts.full=apply(cpts.full(object),1,function(x){sum(x>0,na.rm=TRUE)})
row=which(ncpts.full==ncpts)
if(length(row)==0){
stop(paste("Your input object doesn't have a segmentation with the requested number of changepoints.\n Possible ncpts are: "),paste(ncpts.full,collapse=','))
}
cpts=c(0,cpts.full(object)[row,1:ncpts],length(data.set(object)))
pen.value=pen.value.full(object)[row]
}
nseg=length(cpts)-1
if(test.stat(object)=="Normal"){
if(cpttype(object)=="mean"){
mll.mean=function(x2,x,n){
return( x2-(x^2)/n)
}
y2=c(0,cumsum(data.set(object)^2))
y=c(0,cumsum(data.set(object)))
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.mean(y2[cpts[j+1]+1]-y2[cpts[j]+1],y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
##c(tmplike, tmplike+(nseg-2)*pen.value(object)+sum(log(cpts[-1]-cpts[-(nseg+1)])))
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
else if(cpttype(object)=="variance"){
mll.var=function(x,n){
neg=x<=0
x[neg==TRUE]=0.00000000001
return( n*(log(2*pi)+log(x/n)+1))
}
y2=c(0,cumsum((data.set(object)-param.est(object)$mean)^2))
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.var(y2[cpts[j+1]+1]-y2[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
else if(cpttype(object)=="mean and variance"){
mll.meanvar=function(x2,x,n){
sigmasq=(1/n)*(x2-(x^2)/n)
neg=sigmasq<=0
sigmasq[neg==TRUE]=0.00000000001
return( n*(log(2*pi)+log(sigmasq)+1))
}
y2=c(0,cumsum(data.set(object)^2))
y=c(0,cumsum(data.set(object)))
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvar(y2[cpts[j+1]+1]-y2[cpts[j]+1],y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance' or 'mean and variance'")
}
}
else if(test.stat(object)=="Gamma"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Gamma', must be 'mean and variance'")
}
else{
mll.meanvarg=function(x,n,shape){
return(n*shape*log(n*shape)-n*shape*log(x))
}
y=c(0,cumsum(data.set(object)))
shape=param.est(object)$shape
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvarg(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j],shape)
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
}
else if(test.stat(object)=="Exponential"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Exponential', must be 'mean and variance'")
}
else{
mll.meanvare=function(x,n){
return(n*log(n)-n*log(x))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvare(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
}
else if(test.stat(object)=="Poisson"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Poisson', must be 'mean and variance'")
}
else{
mll.meanvarp=function(x,n){
return(x*log(x)-x*log(n))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvarp(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
}
else{stop("logLik is only valid for distributional assumptions, not CUSUM or CSS")}
return(like)
})
setMethod("logLik", "cpt.reg", function(object) {
if(length(param.est(object))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(object)
cat('done.\n')
}
if(test.stat(object)=="Normal"){
cpts=c(0,object@cpts)
seglen=seg.len(object)
data=data.set(object)
beta=param.est(object)$beta
sigmas=param.est(object)$sig2
rss=NULL
for(i in 1:length(seglen)){
rss[i]=sum((data[(cpts[i]+1):cpts[i+1],1]-data[(cpts[i]+1):cpts[i+1],-1]%*%beta[i,])^2)
}
like=sum(seglen*log(2*pi*sigmas))+sum(rss/sigmas)
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else{stop("logLik is only valid for Normal distributional assumption.")}
return(like)
})
setGeneric("likelihood", function(object) standardGeneric("likelihood"))
setMethod("likelihood", "cpt", function(object) {
return(logLik(object))
})
# acf functions
setGeneric("acf", function(object,...) standardGeneric("acf"))
setMethod("acf", "cpt", function(object,lag.max=NULL,...) {
cpts=c(0,object@cpts)
nseg=nseg(object)
data=data.set(object)
for(i in 1:nseg){
stats::acf(data[(cpts[i]+1):cpts[i+1]],main=paste("Series part:",(cpts[i]+1),":",cpts[i+1]),...)
}
})
setMethod("acf", "cpt.reg", function(object,lag.max=NULL,...) {
cpts=c(0,object@cpts)
nseg=nseg(object)
data=data.set(object)[,1]
for(i in 1:nseg){
stats::acf(data[(cpts[i]+1):cpts[i+1]],main=paste("Series part:",(cpts[i]+1),"-",cpts[i+1]),...)
}
})
Try the changepoint package in your browser
Any scripts or data that you put into this service are public.
changepoint documentation built on Nov. 4, 2022, 1:07 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
setClass("cpt",slots=list(data.set="ts", cpttype="character", method="character", test.stat="character",pen.type="character",pen.value="numeric",minseglen="numeric",cpts="numeric",ncpts.max="numeric",param.est="list",date="character",version="character"),prototype=prototype(cpttype="Not Set",date=date(),version=as(packageVersion("changepoint"),'character')))
setClass("cpt.reg",slots=list(data.set="matrix", cpttype="character", method="character", test.stat="character",pen.type="character",pen.value="numeric",minseglen="numeric",cpts="numeric",ncpts.max="numeric",param.est="list",date="character",version="character"),prototype=prototype(cpttype="regression",date=date(),version=as(packageVersion("changepoint"),"character")))
# setClass("cpt", representation(), prototype())
# # cpts is the optimal segementation
#
setClass("cpt.range",slots=list(cpts.full="matrix", pen.value.full="numeric"), prototype=prototype(), contains="cpt")
# cpts.full is the entire matrix
# pen.value.full (beta) values as an extra slot (vector)
# retrival functions for slots
if(!isGeneric("data.set")) {
if (is.function("data.set")){
fun <- data.set
}
else {fun <- function(object){
standardGeneric("data.set")
}
}
setGeneric("data.set", fun)
}
setMethod("data.set","cpt",function(object) coredata(object@data.set))
setMethod("data.set","cpt.reg",function(object) coredata(object@data.set))
if(!isGeneric("data.set.ts")) {
if (is.function("data.set.ts")){
fun <- data.set.ts
}
else {fun <- function(object){
standardGeneric("data.set.ts")
}
}
setGeneric("data.set.ts", fun)
}
setMethod("data.set.ts","cpt",function(object) object@data.set)
if(!isGeneric("cpttype")) {
if (is.function("cpttype")){
fun <- cpttype
}
else {fun <- function(object){
standardGeneric("cpttype")
}
}
setGeneric("cpttype", fun)
}
setMethod("cpttype","cpt",function(object) object@cpttype)
setMethod("cpttype","cpt.reg",function(object) object@cpttype)
if(!isGeneric("method")) {
if (is.function("method")){
fun <- method
}
else {fun <- function(object){
standardGeneric("method")
}
}
setGeneric("method", fun)
}
setMethod("method","cpt",function(object) object@method)
setMethod("method","cpt.reg",function(object) object@method)
# distribution remains for backwards compatability, changed to test.stat version 1.0
if(!isGeneric("distribution")) {
if (is.function("distribution")){
fun <- distribution
}
else {fun <- function(object){
standardGeneric("distribution")
}
}
setGeneric("distribution", fun)
}
setMethod("distribution","cpt",function(object) object@test.stat)
setMethod("distribution","cpt.reg",function(object) object@test.stat)
if(!isGeneric("test.stat")) {
if (is.function("test.stat")){
fun <- test.stat
}
else {fun <- function(object){
standardGeneric("test.stat")
}
}
setGeneric("test.stat", fun)
}
setMethod("test.stat","cpt",function(object) object@test.stat)
setMethod("test.stat","cpt.reg",function(object) object@test.stat)
if(!isGeneric("pen.type")) {
if (is.function("pen.type")){
fun <- pen.type
}
else {fun <- function(object){
standardGeneric("pen.type")
}
}
setGeneric("pen.type", fun)
}
setMethod("pen.type","cpt",function(object) object@pen.type)
setMethod("pen.type","cpt.reg",function(object) object@pen.type)
if(!isGeneric("pen.value")) {
if (is.function("pen.value")){
fun <- pen.value
}
else {fun <- function(object){
standardGeneric("pen.value")
}
}
setGeneric("pen.value", fun)
}
setMethod("pen.value","cpt",function(object) object@pen.value)
setMethod("pen.value","cpt.reg",function(object) object@pen.value)
if(!isGeneric("pen.value.full")) {
if (is.function("pen.value.full")){
fun <- pen.value.full
}
else {fun <- function(object){
standardGeneric("pen.value.full")
}
}
setGeneric("pen.value.full", fun)
}
setMethod("pen.value.full","cpt.range",function(object) object@pen.value.full)
if(!isGeneric("minseglen")) {
if (is.function("minseglen")){
fun <- minseglen
}
else {fun <- function(object){
standardGeneric("minseglen")
}
}
setGeneric("minseglen", fun)
}
setMethod("minseglen","cpt",function(object) object@minseglen)
if(!isGeneric("cpts")) {
if (is.function("cpts")){
fun <- cpts
}
else {fun <- function(object,...){
standardGeneric("cpts")
}
}
setGeneric("cpts", fun)
}
setMethod("cpts","cpt",function(object) object@cpts[-length(object@cpts)])
setMethod("cpts","cpt.reg",function(object) object@cpts[-length(object@cpts)])
setMethod("cpts","cpt.range",function(object,ncpts=NA){
if(is.na(ncpts)){return(object@cpts[-length(object@cpts)])}
else{
ncpts.full=apply(cpts.full(object),1,function(x){sum(x>0,na.rm=TRUE)})
row=try(which(ncpts.full==ncpts),silent=TRUE)
if(inherits(row,'try-error')){
stop("Your input object doesn't have a segmentation with the requested number of changepoints.\n Possible ncpts are: ",paste(ncpts.full,collapse=','))
}
else{return(cpts.full(object)[row,])}
}
})
if(!isGeneric("cpts.full")) {
if (is.function("cpts.full")){
fun <- cpts.full
}
else {fun <- function(object){
standardGeneric("cpts.full")
}
}
setGeneric("cpts.full", fun)
}
setMethod("cpts.full","cpt.range",function(object) object@cpts.full)
if(!isGeneric("cpts.ts")) {
if (is.function("cpts.ts")){
fun <- cpts.ts
}
else {fun <- function(object){
standardGeneric("cpts.ts")
}
}
setGeneric("cpts.ts", fun)
}
setMethod("cpts.ts","cpt",function(object) index(data.set.ts(object))[cpts(object)] )
if(!isGeneric("ncpts.max")) {
if (is.function("ncpts.max")){
fun <- ncpts.max
}
else {fun <- function(object){
standardGeneric("ncpts.max")
}
}
setGeneric("ncpts.max", fun)
}
setMethod("ncpts.max","cpt",function(object) object@ncpts.max)
setMethod("ncpts.max","cpt.reg",function(object) object@ncpts.max)
if(!isGeneric("param.est")) {
if (is.function("param.est")){
fun <- param.est
}
else {fun <- function(object){
standardGeneric("param.est")
}
}
setGeneric("param.est", fun)
}
setMethod("param.est","cpt",function(object) object@param.est)
setMethod("param.est","cpt.reg",function(object) object@param.est)
setMethod("coef","cpt",function(object) object@param.est)
setMethod("coef","cpt.reg",function(object) object@param.est)
# ncpts function
if(!isGeneric("ncpts")) {
if (is.function("ncpts")){
fun <- ncpts
}
else {fun <- function(object){
standardGeneric("ncpts")
}
}
setGeneric("ncpts", fun)
}
setMethod("ncpts","cpt",function(object) length(cpts(object)))
setMethod("ncpts","cpt.reg",function(object) length(cpts(object)))
# seg.len function
if(!isGeneric("seg.len")) {
if (is.function("seg.len")){
fun <- seg.len
}
else {fun <- function(object){
standardGeneric("seg.len")
}
}
setGeneric("seg.len", fun)
}
setMethod("seg.len","cpt",function(object){object@cpts-c(0,object@cpts[-length(object@cpts)])})
setMethod("seg.len","cpt.reg",function(object){object@cpts-c(0,object@cpts[-length(object@cpts)])})
#i.e. if there is a changepoint in the data, return segment length. If not, return length of the data
# nseg function
if(!isGeneric("nseg")) {
if (is.function("nseg")){
fun <- nseg
}
else {fun <- function(object){
standardGeneric("nseg")
}
}
setGeneric("nseg", fun)
}
setMethod("nseg","cpt",function(object){ncpts(object)+1})
setMethod("nseg","cpt.reg",function(object){ncpts(object)+1})
# replacement functions for slots
setGeneric("data.set<-", function(object, value) standardGeneric("data.set<-"))
setReplaceMethod("data.set", "cpt", function(object, value) {
if(is.ts(value)){object@data.set <- value}else{object@data.set <- ts(value)}
return(object)
})
setReplaceMethod("data.set", "cpt.reg", function(object, value) {
object@data.set <- value
return(object)
})
setGeneric("cpttype<-", function(object, value) standardGeneric("cpttype<-"))
setReplaceMethod("cpttype", "cpt", function(object, value) {
object@cpttype <- value
return(object)
})
setReplaceMethod("cpttype", "cpt.reg", function(object, value) {
object@cpttype <- value
return(object)
})
setGeneric("method<-", function(object, value) standardGeneric("method<-"))
setReplaceMethod("method", "cpt", function(object, value) {
object@method <- value
return(object)
})
setReplaceMethod("method", "cpt.reg", function(object, value) {
object@method <- value
return(object)
})
# distribution remains for backwards compatability, changed to test.stat version 1.0
setGeneric("distribution<-", function(object, value) standardGeneric("distribution<-"))
setReplaceMethod("distribution", "cpt", function(object, value) {
object@test.stat <- value
return(object)
})
setReplaceMethod("distribution", "cpt.reg", function(object, value) {
object@test.stat <- value
return(object)
})
setGeneric("test.stat<-", function(object, value) standardGeneric("test.stat<-"))
setReplaceMethod("test.stat", "cpt", function(object, value) {
object@test.stat <- value
return(object)
})
setReplaceMethod("test.stat", "cpt.reg", function(object, value) {
object@test.stat <- value
return(object)
})
setGeneric("pen.type<-", function(object, value) standardGeneric("pen.type<-"))
setReplaceMethod("pen.type", "cpt", function(object, value) {
object@pen.type <- value
return(object)
})
setReplaceMethod("pen.type", "cpt.reg", function(object, value) {
object@pen.type <- value
return(object)
})
setGeneric("pen.value<-", function(object, value) standardGeneric("pen.value<-"))
setReplaceMethod("pen.value", "cpt", function(object, value) {
object@pen.value <- value
return(object)
})
setReplaceMethod("pen.value", "cpt.reg", function(object, value) {
object@pen.value <- value
return(object)
})
setGeneric("minseglen<-", function(object, value) standardGeneric("minseglen<-"))
setReplaceMethod("minseglen", "cpt", function(object, value) {
object@minseglen <- value
return(object)
})
setReplaceMethod("minseglen", "cpt.range", function(object, value) {
object@minseglen <- value
return(object)
})
setReplaceMethod("minseglen", "cpt.reg", function(object, value) {
object@minseglen <- value
return(object)
})
setGeneric("cpts<-", function(object, value) standardGeneric("cpts<-"))
setReplaceMethod("cpts", "cpt", function(object, value) {
if((cpttype(object)=="meanar")|(cpttype(object)=="trendar")){
n=length(object@data.set)-1
}
else{n=length(object@data.set)}
if(value[length(value)]==n){object@cpts <- value}
else{ object@cpts <- c(value,n) }
return(object)
})
setReplaceMethod("cpts", "cpt.range", function(object, value) {
if((cpttype(object)=="meanar")|(cpttype(object)=="trendar")){
n=length(object@data.set)-1
}
else{n=length(object@data.set)}
if(value[length(value)]==n){object@cpts <- value}
else{ object@cpts <- c(value,n) }
return(object)
})
setReplaceMethod("cpts", "cpt.reg", function(object, value) {
if(value[length(value)]==nrow(object@data.set)){object@cpts <- value}
else{ object@cpts <- c(value,nrow(object@data.set)) }
return(object)
})
setGeneric("ncpts.max<-", function(object, value) standardGeneric("ncpts.max<-"))
setReplaceMethod("ncpts.max", "cpt", function(object, value) {
object@ncpts.max <- value
return(object)
})
setReplaceMethod("ncpts.max", "cpt.reg", function(object, value) {
object@ncpts.max <- value
return(object)
})
setGeneric("param.est<-", function(object, value) standardGeneric("param.est<-"))
setReplaceMethod("param.est", "cpt", function(object, value) {
object@param.est <- value
return(object)
})
setReplaceMethod("param.est", "cpt.reg", function(object, value) {
object@param.est <- value
return(object)
})
setGeneric("cpts.full<-", function(object, value) standardGeneric("cpts.full<-"))
setReplaceMethod("cpts.full", "cpt.range", function(object, value) {
object@cpts.full <- value
return(object)
})
setGeneric("pen.value.full<-", function(object, value) standardGeneric("pen.value.full<-"))
setReplaceMethod("pen.value.full", "cpt.range", function(object, value) {
object@pen.value.full <- value
return(object)
})
# setGeneric("pen.value.input<-", function(object, value) standardGeneric("pen.value.input<-"))
# setReplaceMethod("pen.value.input", "cpt", function(object, value) {
# object@pen.value.input <- value
# return(object)
# })
# parameter functions
setGeneric("param", function(object,...) standardGeneric("param"))
setMethod("param", "cpt", function(object,shape,...) {
param.mean=function(object){
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
data=data.set(object)
tmpmean=NULL
for(j in 1:nseg(object)){
tmpmean[j]=mean(data[(cpts[j]+1):(cpts[j+1])])
}
return(tmpmean)
}
param.var=function(object){
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
data=data.set(object)
seglen=seg.len(object)
tmpvar=NULL
for(j in 1:nseg(object)){
tmpvar[j]=var(data[(cpts[j]+1):(cpts[j+1])])
}
tmpvar=tmpvar*(seglen-1)/seglen # correctly for the fact that the MLE estimate is /n but the var function is /n-1
return(tmpvar)
}
param.scale=function(object,shape){
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
data=data.set(object)
y=c(0,cumsum(data))
tmpscale=NULL
for(j in 1:nseg(object)){
tmpscale[j]=(y[(cpts[j+1]+1)]-y[(cpts[j]+1)])/((cpts[j+1]-cpts[j])*shape)
}
return(tmpscale)
}
param.trend=function(object){
cpts=c(0,object@cpts)
seglen=seg.len(object)
data=data.set(object)
n=length(data)
sumstat=cbind(cumsum(c(0,data)),cumsum(c(0,data*c(1:n))))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=2)
cptsumstat[,2]=cptsumstat[,2]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx3
thetaS=(2*cptsumstat[,1]*(2*seglen + 1) - 6*cptsumstat[,2]) / (2*seglen*(2*seglen + 1) - 3*seglen*(seglen+1))
thetaT=(6*cptsumstat[,2])/((seglen+1)*(2*seglen+1)) + (thetaS * (1-((3*seglen)/((2*seglen)+1))))
return(cbind(thetaS,thetaT))
}
param.meanar=function(object){
seglen=seg.len(object)
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]^2)),cumsum(c(0,data[-(n+1)]^2)))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=5)
beta2=(2*seglen*cptsumstat[,3]-cptsumstat[,1]*cptsumstat[,2])/(2*seglen*cptsumstat[,5]*(1-cptsumstat[,2]^2));
beta1=(2*cptsumstat[,1]-beta2*cptsumstat[,2])/(2*seglen);
return(cbind(beta1,beta2))
}
param.trendar=function(object){
seglen=seg.len(object)
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]*c(1:n))),cumsum(c(0,data[-(n+1)]*c(0:(n-1)))),cumsum(c(0,data[-1]^2)),cumsum(c(0,data[-(n+1)]^2)))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=7)
cptsumstat[,4]=cptsumstat[,4]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx4
cptsumstat[,5]=cptsumstat[,5]-cptsumstat[,2]*c(0,cpts(object)) # i.e. creating newx5
betatop=seglen*(seglen-1)*(seglen*(seglen-1)*cptsumstat[,3] + 2*(2*seglen+1)*cptsumstat[,1]*(cptsumstat[,5]-seglen*cptsumstat[,2]) + 6*cptsumstat[,4]*(cptsumstat[,2]-cptsumstat[,5]))
betabottom=seglen*(seglen-1)*cptsumstat[,7] + 2*(2*seglen+1)*cptsumstat[,2]*(seglen*cptsumstat[,2]-cptsumstat[,5]) + 6*cptsumstat[,5]*(cptsumstat[,5]-cptsumstat[,2]);
beta=betatop/betabottom;
thetajpo=(6*(seglen+2)*(cptsumstat[,4]-beta*cptsumstat[,5]))/((seglen+1)*(2*seglen+1)) - 2*(cptsumstat[,1]-beta*cptsumstat[,2])
thetaj=(2*(2*seglen+1)*(cptsumstat[,1]-beta*cptsumstat[,2])-6*(cptsumstat[,4]-beta*cptsumstat[,5]))/(seglen-1)
return(cbind(beta,thetajpo,thetaj))
}
if(cpttype(object)=="mean"){
param.est(object)<-list(mean=param.mean(object))
}
else if(cpttype(object)=="variance"){
param.est(object)<-list(variance=param.var(object))
}
else if(cpttype(object)=="mean and variance"){
if(test.stat(object)=="Normal"){
param.est(object)<-list(mean=param.mean(object),variance=param.var(object))
}
else if(test.stat(object)=="Gamma"){
param.est(object)<-list(scale=param.scale(object,shape=shape),shape=shape)
}
else if(test.stat(object)=="Exponential"){
param.est(object)<-list(rate=1/param.mean(object))
}
else if(test.stat(object)=="Poisson"){
param.est(object)<-list(lambda=param.mean(object))
}
else{
stop("Unknown test statistic for a change in mean and variance")
}
}
else if(cpttype(object)=="trend"){
if(test.stat(object)=="Normal"){
tmp=param.trend(object)
param.est(object)<-list(thetaS=tmp[,1],thetaT=tmp[,2])
}
else{
stop("Unknown test statistic for a change in trend")
}
}
else if(cpttype(object)=="trendar"){
if(test.stat(object)=="Normal"){
tmp=param.trendar(object)
param.est(object)<-list(beta=tmp[,1],thetajpo=tmp[,2],thetaj=tmp[,3])
}
else{
stop("Unknown test statistic for a change in trend+ar")
}
}
else if(cpttype(object)=="meanar"){
if(test.stat(object)=="Normal"){
tmp=param.meanar(object)
param.est(object)<-list(beta1=tmp[,1],beta2=tmp[,2])
}
else{
stop("Unknown test statistic for a change in mean+ar")
}
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance', 'mean and variance', 'trend', 'meanar' or 'trendar'.")
}
return(object)
})
setMethod("param", "cpt.range", function(object,ncpts=NA,shape,...) {
if(is.na(ncpts)){
cpts=object@cpts
if(cpts[1]!=0){cpts=c(0,cpts)} # PELT derivatives don't include the 0
}
else{
ncpts.full=apply(cpts.full(object),1,function(x){sum(x>0,na.rm=TRUE)})
row=try(which(ncpts.full==ncpts),silent=TRUE)
if(inherits(row,'try-error')){
stop("Your input object doesn't have a segmentation with the requested number of changepoints.")
}
cpts=c(0,cpts.full(object)[row,1:ncpts],length(data.set(object)))
}
param.mean=function(object,cpts){
nseg=length(cpts)-1
data=data.set(object)
tmpmean=NULL
for(j in 1:nseg){
tmpmean[j]=mean(data[(cpts[j]+1):(cpts[j+1])])
}
return(tmpmean)
}
param.var=function(object,cpts){
nseg=length(cpts)-1
data=data.set(object)
seglen=cpts[-1]-cpts[-length(cpts)]
tmpvar=NULL
for(j in 1:nseg){
tmpvar[j]=var(data[(cpts[j]+1):(cpts[j+1])])
}
tmpvar=tmpvar*(seglen-1)/seglen
return(tmpvar)
}
param.scale=function(object,cpts,shape){
nseg=length(cpts)-1
data=data.set(object)
y=c(0,cumsum(data))
tmpscale=NULL
for(j in 1:nseg){
tmpscale[j]=(y[(cpts[j+1]+1)]-y[(cpts[j]+1)])/((cpts[j+1]-cpts[j])*shape)
}
return(tmpscale)
}
param.trend=function(object){
cpts=c(0,object@cpts)
seglen=seg.len(object)
data=data.set(object)
n=length(data)
sumstat=cbind(cumsum(c(0,data)),cumsum(c(0,data*c(1:n))))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=2)
cptsumstat[,2]=cptsumstat[,2]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx3
thetaS=(2*cptsumstat[,1]*(2*seglen + 1) - 6*cptsumstat[,2]) / (2*seglen*(2*seglen + 1) - 3*seglen*(seglen+1))
thetaT=(6*cptsumstat[,2])/((seglen+1)*(2*seglen+1)) + (thetaS * (1-((3*seglen)/((2*seglen)+1))))
return(cbind(thetaS,thetaT))
}
param.meanar=function(object,cpts){
seglen=cpts[-1]-cpts[-length(cpts)]
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]^2)),cumsum(c(0,data[-(n+1)]^2)))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=5)
beta2=(2*seglen*cptsumstat[,3]-cptsumstat[,1]*cptsumstat[,2])/(2*seglen*cptsumstat[,5]*(1-cptsumstat[,2]^2));
beta1=(2*cptsumstat[,1]-beta2*cptsumstat[,2])/(2*seglen);
return(cbind(beta1,beta2))
}
param.trendar=function(object,cpts){
seglen=cpts[-1]-cpts[-length(cpts)]
data=data.set(object)
n=length(data)-1
sumstat=cbind(cumsum(c(0,data[-1])),cumsum(c(0,data[-(n+1)])),cumsum(c(0,data[-1]*data[-(n+1)])),cumsum(c(0,data[-1]*c(1:n))),cumsum(c(0,data[-(n+1)]*c(0:(n-1)))))
cptsumstat=matrix(sumstat[object@cpts+1,]-sumstat[c(0,cpts(object))+1,],ncol=7)
cptsumstat[,4]=cptsumstat[,4]-cptsumstat[,1]*c(0,cpts(object)) # i.e. creating newx4
cptsumstat[,5]=cptsumstat[,5]-cptsumstat[,2]*c(0,cpts(object)) # i.e. creating newx5
betatop=seglen*(seglen-1)*(seglen*(seglen-1)*cptsumstat[,3] + 2*(2*seglen+1)*cptsumstat[,1]*(cptsumstat[,5]-seglen*cptsumstat[,2]) + 6*cptsumstat[,4]*(cptsumstat[,2]-cptsumstat[,5]))
betabottom=seglen*(seglen-1)*cptsumstat[,7] + 2*(2*seglen+1)*cptsumstat[,2]*(seglen*cptsumstat[,2]-cptsumstat[,5]) + 6*cptsumstat[,5]*(cptsumstat[,5]-cptsumstat[,2]);
beta=betatop/betabottom;
thetajpo=(6*(seglen+2)*(cptsumstat[,4]-beta*cptsumstat[,5]))/((seglen+1)*(2*seglen+1)) - 2*(cptsumstat[,1]-beta*cptsumstat[,2])
thetaj=(2*(2*seglen+1)*(cptsumstat[,1]-beta*cptsumstat[,2])-6*(cptsumstat[,4]-beta*cptsumstat[,5]))/(seglen-1)
return(cbind(beta,thetajpo,thetaj))
}
if(cpttype(object)=="mean"){
param.est<-list(mean=param.mean(object,cpts))
}
else if(cpttype(object)=="variance"){
param.est<-list(variance=param.var(object,cpts))
}
else if(cpttype(object)=="mean and variance"){
if(test.stat(object)=="Normal"){
param.est<-list(mean=param.mean(object,cpts),variance=param.var(object,cpts))
}
else if(test.stat(object)=="Gamma"){
param.est<-list(scale=param.scale(object,cpts,shape=shape),shape=shape)
}
else if(test.stat(object)=="Exponential"){
param.est<-list(rate=1/param.mean(object,cpts))
}
else if(test.stat(object)=="Poisson"){
param.est<-list(lambda=param.mean(object,cpts))
}
else{
stop("Unknown test statistic for a change in mean and variance")
}
}
else if(cpttype(object)=="trend"){
if(test.stat(object)=="Normal"){
tmp=param.trend(object)
param.est(object)<-list(thetaS=tmp[,1],thetaT=tmp[,2])
}
else{
stop("Unknown test statistic for a change in trend")
}
}
else if(cpttype(object)=="trendar"){
if(test.stat(object)=="Normal"){
tmp=param.trendar(object)
param.est(object)<-list(beta=tmp[,1],thetajpo=tmp[,2],thetaj=tmp[,3])
}
else{
stop("Unknown test statistic for a change in trend+ar")
}
}
else if(cpttype(object)=="meanar"){
if(test.stat(object)=="Normal"){
tmp=param.meanar(object)
param.est(object)<-list(beta1=tmp[,1],beta2=tmp[,2])
}
else{
stop("Unknown test statistic for a change in mean+ar")
}
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance', 'mean and variance', 'trend', 'meanar' or 'trendar'")
}
if(is.na(ncpts)){
param.est(object)=param.est
return(object)
}
out=new('cpt.range')
param.est(out)=param.est
return(out)
})
setMethod("param", "cpt.reg", function(object,shape,...) {
param.norm=function(object){
cpts=c(0,object@cpts)
# nseg=length(cpts)-1 #nseg(object)
data=data.set(object)
p=ncol(data)-1
tmpbeta=matrix(NA,ncol=p,nrow=nseg(object))
tmpsigma=rep(NA,nseg(object))
for(j in 1:nseg(object)){
formula=paste('-1+data[',cpts[j]+1,':',cpts[j+1],',2]',sep='')
if(p>1){
for(i in 2:p){
formula=paste(formula,'+data[',(cpts[j]+1),':',cpts[j+1],',',i+1,']',sep='')
}
}
tmpfit=eval(parse(text=paste('lm(data[',(cpts[j]+1),':',cpts[j+1],',1]~',formula,')',sep='')))
tmpbeta[j,]=tmpfit$coefficients
tmpsigma[j]=sum(tmpfit$residuals^2)/(length(tmpfit$residuals)-length(tmpfit$coefficients)) ##var(tmpfit$residuals)
}
return(list(beta=tmpbeta,sig2=tmpsigma))
}
if(test.stat(object)=="Normal"){
param.est(object)<-param.norm(object)
}
else{
stop("Unknown test statistic, must be 'Normal'")
}
return(object)
})
# summary functions
setMethod("summary","cpt",function(object){
cat("Created Using changepoint version",object@version,'\n')
cat("Changepoint type : Change in",cpttype(object),'\n')
cat("Method of analysis :",method(object),"\n")
cat("Test Statistic :", test.stat(object),"\n")
cat("Type of penalty :", pen.type(object), "with value,",pen.value(object),"\n")
cat("Minimum Segment Length :", minseglen(object),"\n")
cat("Maximum no. of cpts :", ncpts.max(object),"\n")
if(length(cpts(object))<=20){cat("Changepoint Locations :",cpts(object),"\n")}
else{cat("Number of changepoints:", ncpts(object),"\n")}
})
setMethod("summary","cpt.range",function(object){
cat("Created Using changepoint version",object@version,'\n')
cat("Changepoint type : Change in",cpttype(object),'\n')
cat("Method of analysis :",method(object),"\n")
cat("Test Statistic :", test.stat(object),"\n")
cat("Type of penalty :", pen.type(object), "with value,",pen.value(object),"\n")
cat("Minimum Segment Length :", minseglen(object),"\n")
cat("Maximum no. of cpts :", ncpts.max(object),"\n")
if(length(cpts(object))<=20){cat("Changepoint Locations :",cpts(object),"\n")}
else{cat("Number of changepoints:", ncpts(object),"\n")}
if((nrow(cpts.full(object))<=5)&(ncol(cpts.full(object)<=20))){cat("Range of segmentations:\n");print(cpts.full(object));cat("\n For penalty values:", pen.value.full(object),"\n")}
else{cat("Number of segmentations recorded:", nrow(cpts.full(object)), " with between ", sum(cpts.full(object)[nrow(cpts.full(object)),]>0,na.rm=T), " and ", sum(cpts.full(object)[1,]>0,na.rm=T), "changepoints.\n Penalty value ranges from:",min(pen.value.full(object))," to ",max(pen.value.full(object)))}
})
setMethod("summary","cpt.reg",function(object){
cat("Created Using changepoint version",object@version,'\n')
cat("Changepoint type : Change in",cpttype(object),'\n')
cat("Method of analysis :",method(object),"\n")
cat("Test Statistic :", test.stat(object),"\n")
cat("Type of penalty :", pen.type(object), "with value,",pen.value(object),"\n")
cat("Maximum no. of cpts :", ncpts.max(object),"\n")
if(length(cpts(object))<=20){cat("Changepoint Locations :",cpts(object),"\n")}
else{cat("Number of changepoints:", ncpts(object),"\n")}
})
# show functions
setMethod("show","cpt",function(object){
cat("Class 'cpt' : Changepoint Object\n")
cat(" ~~ : S4 class containing", length(attributes(object))-1, "slots with names\n")
cat(" ", names(attributes(object))[1:(length(attributes(object))-1)], "\n\n")
cat("Created on :", object@date, "\n\n")
cat("summary(.) :\n----------\n")
summary(object)
})
setMethod("show","cpt.reg",function(object){
cat("Class 'cpt.reg' : Changepoint Regression Object\n")
cat(" ~~ : S4 class containing", length(attributes(object))-1, "slots with names\n")
cat(" ", names(attributes(object))[1:(length(attributes(object))-1)], "\n\n")
cat("Created on :", object@date, "\n\n")
cat("summary(.) :\n----------\n")
summary(object)
})
# plot functions
setMethod("plot","cpt",function(x,cpt.col='red',cpt.width=1,cpt.style=1,...){
if(length(param.est(x))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(x)
cat('done.\n')
}
plot(data.set.ts(x),...)
if(cpttype(x)=="variance" || cpttype(x)=="nonparametric (empirical_distribution)"){
abline(v=index(data.set.ts(x))[cpts(x)],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
else if(cpttype(x)=="mean" || cpttype(x)=="mean and variance"){
#nseg=length(cpts(x))+1
cpts=c(0,x@cpts)
if((test.stat(x)=="Normal")||(test.stat(x)=="CUSUM")){
means=param.est(x)$mean
}
else if(test.stat(x)=="Gamma"){
means=param.est(x)$scale*param.est(x)$shape
}
else if(test.stat(x)=="Exponential"){
means=1/param.est(x)$rate
}
else if(test.stat(x)=="Poisson"){
means=param.est(x)$lambda
}
else{
stop('Invalid Changepoint test statistic')
}
for(i in 1:nseg(x)){
segments(index(data.set.ts(x))[cpts[i]+1],means[i],index(data.set.ts(x))[cpts[i+1]],means[i],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else if(cpttype(x)=="trend"){
cpts=c(0,x@cpts)
intercept=rep(param.est(x)$thetaS,x@cpts-c(0,cpts(x)))
slope=rep(param.est(x)$thetaT-param.est(x)$thetaS,x@cpts-c(0,cpts(x)))/rep(x@cpts-c(0,cpts(x)),x@cpts-c(0,cpts(x)))
cptn=rep(c(0,cpts(x)),x@cpts-c(0,cpts(x)))
n=length(data.set(x))
means=intercept+slope*((1:n)-cptn)
for(i in 1:nseg(x)){
segments(index(data.set.ts(x))[cpts[i]+1],means[cpts[i]+1],index(data.set.ts(x))[cpts[i+1]],means[cpts[i+1]],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else{
stop('Invalid Changepoint Type for plotting.\n Can only plot mean, variance, mean and variance')
}
})
setMethod("plot","cpt.range",function(x,ncpts=NA,diagnostic=FALSE,cpt.col='red',cpt.width=1,cpt.style=1,...){
if(diagnostic==TRUE){
n.changepoints = apply(cpts.full(x), 1, function(x) sum(x > 0, na.rm = TRUE))
penalty.values = pen.value.full(x)
if (is.null(list(...)$type)) {
# By default, the type of the diagnostic plots is "lines".
plot(x = n.changepoints, y = penalty.values, xlab = 'Number of Changepoints', ylab = 'Penalty Value', type = "l", ...)
} else {
plot(x = n.changepoints, y = penalty.values, xlab = 'Number of Changepoints', ylab = 'Penalty Value', ...)
}
return(invisible(NULL))
}
plot(data.set.ts(x),...)
if(is.na(ncpts)){
if(pen.type(x)=="CROPS"){
stop('CROPS does not supply an optimal set of changepoints, set ncpts to the desired segmentation to plot or use diagnostic=TRUE to identify an appropriate number of changepoints')
}
cpts.to.plot=cpts(x)
param.est=x
}
else{
ncpts.full=apply(cpts.full(x),1,function(x){sum(x>0,na.rm=TRUE)})
row=which(ncpts.full==ncpts)
if(length(row)==0){
stop(paste("Your input object doesn't have a segmentation with the requested number of changepoints.\n Possible ncpts are: "),paste(ncpts.full,collapse=','))
}
cpts.to.plot=cpts.full(x)[row,1:ncpts]
if(test.stat(x)=="Gamma"){
param.est=param(x,ncpts,shape=param.est(x)$shape)
}
else{
param.est=param(x,ncpts)
}
}
if(cpttype(x)=="variance"){
abline(v=index(data.set.ts(x))[cpts.to.plot],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
else if(cpttype(x)=="mean" || cpttype(x)=="mean and variance"){
if((test.stat(x)=="Normal")||(test.stat(x)=="CUSUM")){
means=param.est(param.est)$mean
}
else if(test.stat(x)=="Gamma"){
means=param.est(param.est)$scale*param.est(param.est)$shape
}
else if(test.stat(x)=="Exponential"){
means=1/param.est(param.est)$rate
}
else if(test.stat(x)=="Poisson"){
means=param.est(param.est)$lambda
}
else{
stop('Invalid Changepoint test statistic')
}
nseg=length(means)
cpts.to.plot=c(0,cpts.to.plot,length(data.set(x)))
for(i in 1:nseg){
segments(index(data.set.ts(x))[cpts.to.plot[i]+1],means[i],index(data.set.ts(x))[cpts.to.plot[i+1]],means[i],col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else{
stop('Invalid Changepoint Type for plotting.\n Can only plot mean, variance, mean and variance')
}
})
setMethod("plot","cpt.reg",function(x,cpt.col='red',cpt.width=1,cpt.style=1,...){
if(length(param.est(x))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(x)
cat('done.\n')
}
plot(data.set(x)[,1],type='l',...)
if(test.stat(x)=="Normal"){
cpts=c(0,x@cpts)
betas=param.est(x)$beta
for(i in 1:nseg(x)){
lines((cpts[i]+1):cpts[i+1],betas[i,]%*%t(data.set(x)[(cpts[i]+1):cpts[i+1],-1]),col=cpt.col,lwd=cpt.width,lty=cpt.style)
}
}
else{
stop('Invalid Changepoint test statistic')
}
})
# likelihood functions
setMethod("logLik", "cpt", function(object) {
if(length(param.est(object))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(object)
cat('done.\n')
}
if(test.stat(object)=="Normal"){
if(cpttype(object)=="mean"){
means=rep(param.est(object)$mean,object@cpts-c(0,cpts(object)))
rss=sum((data.set(object)-means)^2)
n=length(data.set(object))
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
cpts=c(0,object@cpts)
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(cpts[-1]-cpts[-(nseg(object)+1)])))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="variance"){
rss=c(0,cumsum((data.set(object)-param.est(object)$mean)^2))
cpts=c(0,object@cpts)
n=length(data.set(object))
seglen=seg.len(object)
sigmas=(rss[cpts[-1]+1]-rss[cpts[-length(cpts)]+1])/seglen
like=n*log(2*pi)+sum(seglen*log(sigmas))+n
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seglen)))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="mean and variance"){
means=rep(param.est(object)$mean,object@cpts-c(0,cpts(object)))
rss=sum((data.set(object)-means)^2)
n=length(data.set(object))
cpts=c(0,object@cpts)
seglen=seg.len(object)
sigmas=param.est(object)$variance
like=n*log(2*pi)+sum(seglen*log(sigmas))+n
if(pen.type(object)=="MBIC"){
like=c(like,like+(nseg(object)-2)*pen.value(object)+sum(log(seglen)))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="trend"){
intercept=rep(param.est(object)$thetaS,object@cpts-c(0,cpts(object)))
slope=rep(param.est(object)$thetaT-param.est(object)$thetaS,object@cpts-c(0,cpts(object)))/rep(object@cpts-c(0,cpts(object)),object@cpts-c(0,cpts(object)))
cptn=rep(c(0,cpts(object)),object@cpts-c(0,cpts(object)))
n=length(data.set(object))
means=intercept+slope*((1:n)-cptn)
rss=sum((data.set(object)-means)^2)
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="trendar"){
seglen=seg.len(object)
intercept=rep(param.est(object)$thetaj,seglen)
slope=rep(param.est(object)$thetajpo-param.est(object)$thetaj,seglen)/rep(seglen,seglen)
ar=rep(param.est(object)$beta,seglen)
cptn=rep(c(0,cpts(object)),seglen)
n=length(data.set(object))
means=NULL;means[1]=0
for(i in 2:n){means[i]=intercept+slope*((1:n)-cptn)+ar*means[i-1]}
means=means[-1]
rss=sum((data.set(object)[-1]-means)^2)
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else if(cpttype(object)=="meanar"){
seglen=seg.len(object)
intercept=rep(param.est(object)$beta1,seglen)
ar=rep(param.est(object)$beta2,seglen)
cptn=rep(c(0,cpts(object)),seglen)
n=length(data.set(object))
means[1]=0;for(i in 2:n){means[i]=intercept+ar*means[i-1]}
means=means[-1]
rss=sum((data.set(object)[-1]-means)^2)
like=n*(log(2*pi)+log(rss/n)+1) # -2*loglik
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance', 'mean and variance', 'trend', 'meanar' or 'trendar'")
}
}
else if(test.stat(object)=="Gamma"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Gamma', must be 'mean and variance'")
}
else{
warning("Not changed to be -2*logLik")
mll.meanvarg=function(x,n,shape){
return(n*shape*log(n*shape)-n*shape*log(x))
}
y=c(0,cumsum(data.set(object)))
shape=param.est(object)$shape
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg(object)){
tmplike=tmplike+mll.meanvarg(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j],shape)
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg(object)-1)*pen.value(object))
}
}
}
else if(test.stat(object)=="Exponential"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Exponential', must be 'mean and variance'")
}
else{
warning("Not changed to be -2*logLik")
mll.meanvare=function(x,n){
return(n*log(n)-n*log(x))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg(object)){
tmplike=tmplike+mll.meanvare(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg(object)-1)*pen.value(object))
}
}
}
else if(test.stat(object)=="Poisson"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Poisson', must be 'mean and variance'")
}
else{
warning("Not changed to be -2*logLik")
mll.meanvarp=function(x,n){
return(x*log(x)-x*log(n))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg(object)){
tmplike=tmplike+mll.meanvarp(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg(object)-1)*pen.value(object))
}
}
}
else{stop("logLik is only valid for distributional assumptions, not CUSUM or CSS")}
names(like)=c("-2*logLik","-2*Loglike+pen")
return(like)
})
setMethod("logLik", "cpt.range", function(object,ncpts=NA) {
warning("Not changed to be -2*logLik")
if(is.na(ncpts)){
if(pen.type(object)=="CROPS"){
stop('CROPS does not supply an optimal set of changepoints, set ncpts argument to the desired segmentation to plot or use diagnostic=TRUE to identify an appropriate number of changepoints')
}
cpts=c(0,object@cpts)
pen.value=pen.value(object)
}
else{
ncpts.full=apply(cpts.full(object),1,function(x){sum(x>0,na.rm=TRUE)})
row=which(ncpts.full==ncpts)
if(length(row)==0){
stop(paste("Your input object doesn't have a segmentation with the requested number of changepoints.\n Possible ncpts are: "),paste(ncpts.full,collapse=','))
}
cpts=c(0,cpts.full(object)[row,1:ncpts],length(data.set(object)))
pen.value=pen.value.full(object)[row]
}
nseg=length(cpts)-1
if(test.stat(object)=="Normal"){
if(cpttype(object)=="mean"){
mll.mean=function(x2,x,n){
return( x2-(x^2)/n)
}
y2=c(0,cumsum(data.set(object)^2))
y=c(0,cumsum(data.set(object)))
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.mean(y2[cpts[j+1]+1]-y2[cpts[j]+1],y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
##c(tmplike, tmplike+(nseg-2)*pen.value(object)+sum(log(cpts[-1]-cpts[-(nseg+1)])))
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
else if(cpttype(object)=="variance"){
mll.var=function(x,n){
neg=x<=0
x[neg==TRUE]=0.00000000001
return( n*(log(2*pi)+log(x/n)+1))
}
y2=c(0,cumsum((data.set(object)-param.est(object)$mean)^2))
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.var(y2[cpts[j+1]+1]-y2[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
else if(cpttype(object)=="mean and variance"){
mll.meanvar=function(x2,x,n){
sigmasq=(1/n)*(x2-(x^2)/n)
neg=sigmasq<=0
sigmasq[neg==TRUE]=0.00000000001
return( n*(log(2*pi)+log(sigmasq)+1))
}
y2=c(0,cumsum(data.set(object)^2))
y=c(0,cumsum(data.set(object)))
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvar(y2[cpts[j+1]+1]-y2[cpts[j]+1],y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
else{
stop("Unknown changepoint type, must be 'mean', 'variance' or 'mean and variance'")
}
}
else if(test.stat(object)=="Gamma"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Gamma', must be 'mean and variance'")
}
else{
mll.meanvarg=function(x,n,shape){
return(n*shape*log(n*shape)-n*shape*log(x))
}
y=c(0,cumsum(data.set(object)))
shape=param.est(object)$shape
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvarg(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j],shape)
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
}
else if(test.stat(object)=="Exponential"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Exponential', must be 'mean and variance'")
}
else{
mll.meanvare=function(x,n){
return(n*log(n)-n*log(x))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvare(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
}
else if(test.stat(object)=="Poisson"){
if(cpttype(object)!="mean and variance"){
stop("Unknown changepoint type for test.stat='Poisson', must be 'mean and variance'")
}
else{
mll.meanvarp=function(x,n){
return(x*log(x)-x*log(n))
}
y=c(0,cumsum(data.set(object)))
cpts=c(0,object@cpts)
#nseg=length(cpts)-1
tmplike=0
for(j in 1:nseg){
tmplike=tmplike+mll.meanvarp(y[cpts[j+1]+1]-y[cpts[j]+1],cpts[j+1]-cpts[j])
}
if(pen.type(object)=="MBIC"){
like=c(tmplike, tmplike+(nseg-2)*pen.value+sum(log(seg.len(object))))
}else{
like=c(tmplike,tmplike+(nseg-1)*pen.value)
}
names(like)=c("-like","-likepen")
}
}
else{stop("logLik is only valid for distributional assumptions, not CUSUM or CSS")}
return(like)
})
setMethod("logLik", "cpt.reg", function(object) {
if(length(param.est(object))==0){# i.e. parameter.estimates=FALSE in call
cat('Calculating parameter estimates...')
object=param(object)
cat('done.\n')
}
if(test.stat(object)=="Normal"){
cpts=c(0,object@cpts)
seglen=seg.len(object)
data=data.set(object)
beta=param.est(object)$beta
sigmas=param.est(object)$sig2
rss=NULL
for(i in 1:length(seglen)){
rss[i]=sum((data[(cpts[i]+1):cpts[i+1],1]-data[(cpts[i]+1):cpts[i+1],-1]%*%beta[i,])^2)
}
like=sum(seglen*log(2*pi*sigmas))+sum(rss/sigmas)
if(pen.type(object)=="MBIC"){
like=c(like, like+(nseg(object)-2)*pen.value(object)+sum(log(seg.len(object))))
}else{
like=c(like,like+(nseg(object)-1)*pen.value(object))
}
}
else{stop("logLik is only valid for Normal distributional assumption.")}
return(like)
})
setGeneric("likelihood", function(object) standardGeneric("likelihood"))
setMethod("likelihood", "cpt", function(object) {
return(logLik(object))
})
# acf functions
setGeneric("acf", function(object,...) standardGeneric("acf"))
setMethod("acf", "cpt", function(object,lag.max=NULL,...) {
cpts=c(0,object@cpts)
nseg=nseg(object)
data=data.set(object)
for(i in 1:nseg){
stats::acf(data[(cpts[i]+1):cpts[i+1]],main=paste("Series part:",(cpts[i]+1),":",cpts[i+1]),...)
}
})
setMethod("acf", "cpt.reg", function(object,lag.max=NULL,...) {
cpts=c(0,object@cpts)
nseg=nseg(object)
data=data.set(object)[,1]
for(i in 1:nseg){
stats::acf(data[(cpts[i]+1):cpts[i+1]],main=paste("Series part:",(cpts[i]+1),"-",cpts[i+1]),...)
}
})
Try the changepoint package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.