ERP package
Outline
data
1. ERPdata : maybe a formal data for package
preprocessing
1. list_to_df
2. data_select
3. data_summarize
4. downsample
EDA
1.edaplot
Data
1. ERPdata
2. NIRSdata
library(ERP);require(mnormt);require(fdrtool);library(ggplot2);library(dplyr)
library(erpR);require(akima);library(reshape2);library(boot)
data(ERPsets)
load("ERPdata.RData")preprocessing
1. list_to_df
2. data_select
3. data_summarize
###### 4. Down sample
list_to_df
It is a function that could transformed a list to a dataframe for our package.
###list2data
list_to_df <- function(list_data,frames){
# some check function
for (i in 1:length(list_data)){
list_data[[i]]$frames <-frames
list_data[[i]]$list.name <- names(list_data)[i]
}
Data_list <- lapply(list_data,melt,id=c("frames","list.name"))
Data_list <- lapply(Data_list,reshape,
timevar = "frames",idvar = c("variable","list.name"),
direction = "wide")
Ana_data <- Data_list[[1]]
for (i in 2:length(ERPsets)){
Ana_data <- rbind(Ana_data,Data_list[[i]])
}
colnames(Ana_data)[2] <- "Channel"
rownames(Ana_data) <- 1:dim(Ana_data)[1]
return(Ana_data)
}ERP_df <- list_to_df(list_data=ERPsets, # input list_data
frames= 1:426) # input frames
ERP_df <- AddExpCondSub(data = ERP_df, # input the transformed data
list.name_col = 1) # list.name column
#ERP_df <- ERP_df[,-1] # remove list name column, Depend on user's need
head(ERP_df[,c(1:3,427:430)],3) # look at the data list.name Channel value.1 value.425 value.426 Experiment
1 Exp1_word_subj1 Fp1 0.36995740 14.160980 9.971283 Exp1
2 Exp1_word_subj1 Fp2 -0.02759907 13.750350 9.678596 Exp1
3 Exp1_word_subj1 F3 0.42971300 6.269233 4.413941 Exp1
Condition
1 word
2 word
3 word2. data_select
It could select data by subject and channel
data_select <- function(data,frames,datacol,subjcol=NULL,chancol=NULL,othvarcol=NULL,
select_subj=NULL,
select_chan=NULL,...){
# some check function
subj_select <- function(data,frames,datacol,subjcol,chancol=NULL,othvarcol=NULL,
select_subj,...){
dta <- data
num <- length(select_subj)
data_list <- list()
for (i in 1:num){
data_new <- subset(dta,dta[,subjcol]==select_subj[i])
data_list[[i]] <- data_new
}
data_select <- data_list[[1]]
if (num != 1) {
for (i in 2:num) {
data_select <- rbind(data_select,data_list[[i]])
}
}
rownames(data_select) <- 1:dim(data_select)[1]
return (data_select)
}
chan_select <- function(data,frames,datacol,subjcol=NULL,chancol,othvarcol=NULL,
select_chan,...){
dta <- data
num <- length(select_chan)
data_list <- list()
for (i in 1:num){
data_new <- subset(dta,dta[,chancol]==select_chan[i])
data_list[[i]] <- data_new
}
data_select <- data_list[[1]]
if (num != 1){
for (i in 2:num){
data_select <- rbind(data_select,data_list[[i]])
}
}
rownames(data_select) <- 1:dim(data_select)[1]
return (data_select)
}
joint_select <- function(data,frames,datacol,subjcol,chancol,othvarcol=NULL,
select_subj,
select_chan,...){
dta <- data
ind_data <- subj_select(dta,frames,datacol,subjcol,chancol,othvarcol,
select_subj = select_subj)
ind_ele_data <- chan_select(ind_data,frames,datacol,subjcol,chancol,othvarcol,
select_chan = select_chan)
rownames(ind_ele_data) <- 1 :dim(ind_ele_data)[1]
return(ind_ele_data)
}
if (is.null(select_subj)==FALSE & is.null(select_chan)==FALSE){
dta <- joint_select(data = data,frames = frames,
datacol,subjcol,chancol,othvarcol,
select_subj=select_subj,
select_chan=select_chan)
} else if (is.null(select_subj)==FALSE & is.null(select_chan)== TRUE) {
dta <- subj_select(data=data,frames=frames,
datacol,subjcol,chancol,othvarcol,
select_subj=select_subj)
} else if (is.null(select_chan)==FALSE & is.null(select_subj)== TRUE) {
dta <- chan_select(data,frames,
datacol,subjcol,chancol,othvarcol,
select_chan=select_chan)
} else {
dta <- data
}
return(dta)
}S1S10Fp1Fp2 <- data_select(data = ERPdata,
frames = 1:426,
datacol = 2:427,
subjcol = 430,
chancol = 1,
othvarcol = c(428:429,431:432),
select_subj = c("subj1","subj2"),
select_chan=c("CZ","Fp1"))
CZ <- data_select(data = ERPdata,
frames = 1:426,
datacol = 2:427,
subjcol = 430,
chancol = 1,
othvarcol = c(428:429,431:432),
#select_subj = c("subj1","subj10"),
select_chan=c("CZ"))
S1S10 <- data_select(data = ERPdata,
frames = 1:426,
datacol = 2:427,
subjcol = 430,
chancol = 1,
othvarcol = c(428:429,431:432),
select_subj = c("subj1","subj10"))
dim(S1S10Fp1Fp2);dim(CZ);dim(S1S10)## [1] 8 432## [1] 40 432## [1] 136 4323. Data_summarize :
We could aggregate data by the column like conditions, subjects or channel.
data_summarize <- function(data,frames,datacol,subjcol=NULL,chancol=NULL,othvarcol=NULL,
summarycol,
fun=mean,
select_subj=NULL,
select_chan=NULL,...){
#some check function
options(warn=-1) # should close the warnings?
# selection
dta <- data_select(data,frames,datacol,subjcol,chancol,othvarcol,
select_subj,
select_chan)
# process
agglength <- length(summarycol)
aggvar_list <- list(dta[,summarycol[1]])
if (agglength > 1){
for (i in 2:agglength ){
aggvar_list <- append(aggvar_list,list(dta[,summarycol[i]]))
}
}
aggdata <- aggregate(dta[,datacol],by=aggvar_list,
fun,...)
aggdata <- aggdata[,1:(agglength+length(datacol))]
for (i in 1: agglength){
colnames(aggdata)[i] <- colnames(dta)[summarycol[i]]
}
rownames(aggdata) <- 1:dim(aggdata)[1]
return(aggdata)
}A <-data_summarize (data = ERPdata,
frames = 1:426,
datacol = 2:427,
#you could put all elements you want to aggregate in datacol argument
subjcol = 430,
chancol = 1,
othvarcol = c(428:429,431:432),
summarycol= c(1,429),
fun=median,
# could also select data
select_chan = c("CZ","Fp1"),
select_subj = c("subj1","subj10"))
tail(A[,1:6]);dim(A) Channel Condition value.1 value.2 value.3 value.4
1 Fp1 nonword -1.4049110 -1.9667935 -1.7905150 -1.7634650
2 CZ nonword -1.8309385 -2.4948925 -2.1159490 -1.9471855
3 Fp1 word 0.4695684 0.1779768 -0.5842498 -0.9740931
4 CZ word 0.8391661 0.8317206 0.2696740 0.1090901[1] 4 4284. Down Sample:
We could down sample the signal
downsample <- function(data,
datacol,
binwidth=10,
movinginterval=NULL) {
if (is.null(movinginterval) == FALSE) {
if (movinginterval >= binwidth){stop("movinginterval should not bigger than binwidth!")}
dta <- data
dta_signal <- dta[,datacol]
dta_othvar <- dta[,-datacol]
num1 <- (dim(dta_signal)[2] - (dim(dta_signal)[2] %% binwidth)) /binwidth
num2 <- dim(dta_signal)[2] %% binwidth
dta_downsample <- data.frame(melt(apply(dta_signal[,1:binwidth],1,mean)))
i = movinginterval - 1
while ((binwidth+i) <= dim(dta_signal)[2]){
dta_downsample <- cbind(dta_downsample,melt(apply(dta_signal[,(1+i):(binwidth+i)],1,mean)))
i = i + movinginterval -1
}
if (num2 >= 1) {
dta_downsample <- cbind(dta_downsample,
value=melt(apply(dta_signal[,(1+i):dim(dta_signal)[2]],1,mean)))
}
colnames(dta_downsample) <- paste("value",1:dim(dta_downsample)[2],sep=".")
dta_final <- cbind(dta_othvar,dta_downsample)
} else {
dta <- data
dta_signal <- dta[,datacol]
dta_othvar <- dta[,-datacol]
num1 <- (dim(dta_signal)[2] - (dim(dta_signal)[2] %% binwidth))/binwidth
num2 <- dim(dta_signal)[2] %% binwidth
dta_downsample <- data.frame(melt(apply(dta_signal[,1:binwidth],1,mean)))
for (i in 1 : num1-1){
dta_downsample <- cbind(dta_downsample,
melt(apply(dta_signal[,(i*binwidth+1):((i+1)*binwidth)],
1,mean)))
}
if (num2 == 1) {
dta_downsample <- cbind(dta_downsample,
value=dta_signal[,(num1*binwidth+num2)])
###### notice
}
if (num2 > 1) {
dta_downsample <- cbind(dta_downsample,
melt(apply(dta_signal[,(num1*binwidth+1):(num1*binwidth+num2)],
1,mean)))
}
dta_downsample <- dta_downsample[,-1]
colnames(dta_downsample) <- paste("value",1:dim(dta_downsample)[2],sep=".")
dta_final <- cbind(dta_othvar,dta_downsample)
}
return(dta_final)
}down1 <- downsample(data = ERPdata,
datacol = 2:427,
binwidth = 10,
movinginterval=NULL)
down2 <- downsample(data = ERPdata,
datacol = 2:427,
binwidth = 10,
movinginterval=9) # movinginterval shoud not bigger than binwidth
dim(down1) ; dim(down2)[1] 1360 49[1] 1360 60EDA
##### 1.edaplot
1.edaplot
The function “edaplot”" uses ggplot2 grammar, so it has several flexible options.
# group comparison please put in data after aggregate_raw
edaplot <- function(data,frames=NULL,datacol,subjcol=NULL,chancol=NULL,othvarcol=NULL,
outlinesub=NULL,outcolor="red",
select_subj=c(NULL),
select_chan=c(NULL),...){
#some check functions
#if (is.null(frames) == FALSE)
# if (length(frames) != (ncol(data)-1-length(othvarcol)))
# stop(paste("frames should be either null or of length",
# (ncol(data)-1-length(othvarcol))))
#if (is.null(frames) == FALSE) {
# if (any(frames != sort(frames)))
# stop("frames should be an ascending sequence of integers")
# }
#if (is.null(frames))
# frames = 1:(ncol(data)-1-length(othvarcol))
#selection
dta <- data_select(data,frames,datacol,subjcol,chancol,othvarcol,
select_subj,
select_chan,...)
# plot
subvar <- variable.names(dta)[subjcol]
dta$groupvar <- rownames(dta)
datalong <- melt(dta,
id=c(variable.names(dta)[c(subjcol,chancol,othvarcol)],
"groupvar"))
datalong <- datalong[order(datalong$groupvar),]
datalongorder <- datalong
datalongorder$frames <- rep(frames,length(datalongorder[,1])/length(frames))
if (is.null(outlinesub) == FALSE){ # how to outline several subjects (and color)
data2 <- subset(datalongorder,datalongorder[,1]==outlinesub)
plot <- ggplot(datalongorder,
aes(x=frames,y=value,group=groupvar,...))+
geom_line()+
geom_line(data=data2,aes(x=frames,y=value),col=outcolor)
# need warning for covering geom_line()
} else {
plot <- ggplot(datalongorder,
aes(x=frames,y=value,group=groupvar))+
geom_line()
}
return(plot)
}#Although Full data (all trials) work fine, I recommened that the argument erpdata should be a single subject data or an aggregate data.
edaplot(ERPdata,
frames = 1:426,
datacol=2:427,
subjcol=430,
chancol=1,
othvarcol=c(428:429,431:432),
# choose the subject you want to show(optional)
select_sub = c("subj9","subj1","subj5","subj8"),
# choose the channel you want to show(optional)
select_chan = c("F3","F4"))+
facet_grid(Channel~Condition)
edaplot(ERPdata,
frames = 1:426,
datacol=2:427,
subjcol=430,
chancol=1,
othvarcol=c(428:429,431:432),
# choose the subject you want to show(optional)
select_sub = c("subj9","subj1","subj5","subj8"),
# choose the channel you want to show(optional)
select_chan = c("F3","F4"),
# highlight a single subject with the color you want
outlinesub="subj9",outcolor = "blue")+
facet_grid(Channel~Condition)
edaplot(ERPdata,
frames=1:426,
datacol = 2:427,subjcol=430,chancol=1,othvarcol=c(428:429,431:432),
select_chan = c("F3","F4","CZ"),
select_sub = c("subj9","subj1","subj5","subj8"))+
# One subject one color
geom_line(aes(col=Subject))+ # will cover outline sub
facet_grid(Condition~Channel)
edaplot(ERPdata,
frames=1:426,
datacol = 2:427,subjcol=430,chancol=1,othvarcol=c(428:429,431:432),
select_chan = c("F3","F4","CZ"))+
# Color depend on IQ (Size , alpha either)
geom_line(aes(col=IQ,alpha=Channel,size=Condition))+
facet_grid(Condition~Channel)
edaplot(ERPdata,
frames=1:426,
datacol = 2:427,subjcol=430,chancol=1,othvarcol=c(428:429,431:432),
select_chan = c("F3","F4","CZ"))+
# One Condition one color
geom_line(aes(col=Condition))+
facet_grid(.~Channel)
edaplot(ERPdata,
frames=1:426,
datacol = 2:427,subjcol=430,chancol=1,othvarcol=c(428:429,431:432),
select_chan = c("F3","F4","CZ"))+
# One Condition one color
geom_line(aes(col=Condition))+
facet_grid(.~Channel)+
scale_colour_manual(values=c("red","blue"),name="Cond",label=c("A","B"))
edaplot(ERPdata,
frames=1:426,
datacol = 2:427,subjcol=430,chancol=1,othvarcol=c(428:429,431:432),
select_chan = c("F3","F4","CZ"))+
geom_line(aes(col=Condition))+
facet_grid(Condition~Channel)
edaplot(ERPdata,
frames=1:426,
datacol = 2:427,subjcol=430,chancol=1,othvarcol=c(428:429,431:432),
select_chan = c("F3","F4","CZ"))+
geom_line(aes(col=Condition))+
facet_grid(Condition~Channel)+
# put on the summary line NOTE: need a group=NULL argument
stat_summary(aes(group=NULL),fun.y = "mean",
colour = "purple", size = 0.5, geom = "line")
edaplot(ERPdata,
frames=1:426,
datacol = 2:427,subjcol=430,chancol=1,othvarcol=c(428:429,431:432),
select_chan = c("F3","F4","CZ"))+
geom_line(aes(col=Condition))+
facet_grid(Condition~Channel)+
theme_bw()+
theme(legend.position="top")+
xlim(-100,500)+
ylim(-50,50)+
labs(list(title = "Flexible",x="time",y="signal"))+
geom_vline(xintercept = c(0,213,426),col="yellow")