1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
|
#' Remove linear trends
#'
#' Removes the mean value or linear trends in each of the input and output matrices.
#'
#' @param data an object of class \code{idframe}
#' @param type trend type - "constant" or "linear". (Default: \code{"linear"})
#'
#' @return
#' A list containing the following elements
#'
#' \tabular{ll}{
#' \code{fitted.values} \tab \code{idframe} object with detrended variables \cr
#' \code{output.trend} \tab \code{list} containing trend fits for each output
#' variable \cr
#' \code{input.trend} \tab \code{list} containing trend fits for each input
#' variable
#' }
#'
#' @examples
#' data(cstr)
#' fit <- detrend(cstr) # remove linear trends
#' Zdetrend <- predict(fit) # get the detrended data
#'
#' demean <- detrend(cstr,type="constant") # remove mean values
#' Zcent <- predict(demean) # get the centered data
#'
#' @seealso \code{\link{predict.detrend}}, \code{\link[stats]{lm}}
#' @export
detrend <- function(data,type=c("constant","linear")[2]){
if(!(type %in% c("constant","linear"))){
stop("Error: Invalid trend type")
}
reg <- time(data$output[,1])
if(type=="linear"){
formula <- X ~ reg
} else {
formula <- X ~ 1 + offset(0*reg)
}
data_detrend <- data
out <- data$output;output_trend <- list()
for(i in 1:ncol(out)){
output_trend[[i]] <- lm(formula,data=data.frame(X=out[,i],reg=reg))
out[,i] <- fitted(output_trend[[i]])
}
input <- data$input;input_trend <- list()
for(i in 1:ncol(input)){
input_trend[[i]] <- lm(formula,data=data.frame(X=input[,i],reg=reg))
input[,i] <- fitted(input_trend[[i]])
}
data_detrend$output <- data$output - out;data_detrend$input <- data$input - input
est <- list(fitted.values=data_detrend,output.trend = output_trend,
input.trend = input_trend)
class(est) <- "detrend"
return(est)
}
#' Predict method for trend fits on idframe objects
#'
#' Detrended \code{idframe} object based on linear trend fit
#'
#' @param object an object of class \code{idframe}
#' @param newdata An optional idframe object in which to look for variables with
#' which to predict. If ommited, the original detrended idframe object is used
#'
#' @return an \code{idframe} object
#'
#' @examples
#' data(cstr)
#' train <- dataSlice(cstr,end=5000) # subset the first 5000 indices
#' test <- dataSlice(cstr,start=6001) # subset from index 6001 till the end
#' fit <- detrend(train)
#' Ztrain <- predict(fit)
#' Ztest <- predict(fit,test)
#'
#' @export
predict.detrend <- function(object,newdata=NULL,...){
if(is.null(newdata)){
data <- fitted(object)
} else{
data <- newdata
out <- detrend.predict(object$output.trend,data$output)
input <- detrend.predict(object$input.trend,data$input)
data$output <- data$output - out
data$input <- data$input - input
}
return(data)
}
detrend.predict <- function(object,data){
pred_list <- list()
for(i in 1:ncol(data)){
pred_list[[i]] <- predict(object[[i]],newdata=data.frame(reg = time(data[,i])))
}
pred <- data.frame(matrix(unlist(pred_list),ncol=ncol(data),byrow=T))
colnames(pred) <- colnames(data)
return(pred)
}
|