R/meanLocationT.R

In ateshkoul/KinematicAnalysis: Generic Package for Kinematic Analysis

#' Function to get mean values of X and Y co-ordinates after time = time for each subject
#'
#' Assumes the rawData has columns = X,Y and NormalisedTime
#' it also has a column Subject
#'
#'@author
#'Atesh Koul, RBCS, Istituto Italiano di technologia
#'
#'\email{atesh.koul@@gmail.com}
meanLocationT <- function(RawData,time,order){
  # Function to get mean values of X and Y co-ordinates after time = time
  # for each subject

  # Assumes the rawData has columns = X,Y and NormalisedTime
  # it also has a column Subject


  # Author:
  # Atesh Koul, RBCS (atesh.koul@iit.it)
  # Istituto Italiano di technologia
  # last changes
  if(!require(plyr)){install.packages('plyr')}

  if(missing(order)){
    order <- "No"
  }



  # libraries

  require(plyr)
  # preprocessing and checking


  # Use only the time after start of the Message
  # 12-05-2015
  # some issues with the way Raw Data was processed.
  # Using CorrectTime was fine but using Normalised Time instead created NA values
  # Don't know why
  RawData <- RawData[RawData$CorrectTime>0,]


  RawData$X <- as.numeric(RawData$X)
  RawData$Y <- as.numeric(RawData$Y)
  RawData$NormalisedTime <- as.numeric(RawData$NormalisedTime)

  MeanSubjects <- list()
  MeanSubjectsTrialHit <- list()
  MeanSubjectsTrialFixation <- list()
  for(i in 1:length(time)){
    # necessary to include an interval as most values would not have exactly the time as .5
    # The value 0.003 is arbitrary (in some sense) as it is derived as a general value from the
    # difference of two consecutive acquisitions in the data

    # Inserting Trial is necessary so as to remove the outliers
    # Currently (12-05-2015) there doesn't seem to be a good way to encode bad Trials in the
    # Raw time series data. It's easier to perform it at the interpretation level
    MeanSubjectsTrialHit[[i]] <- ddply(RawData[which(RawData$NormalisedTime>time[i]-0.006 & RawData$NormalisedTime<time[i]+0.006),],c(paste("Subject.",order,sep=""),"Trial"),summarise,
                               MeanX = mean(X,na.rm=T),MeanY = mean(Y,na.rm=T),Time=NormalisedTime[1])

    # Per Subject does not make a lot of sense because it will mean the locations over multiple trials which have different hand positions
    #     MeanSubjects[[i]] <- ddply(RawData[which(RawData$NormalisedTime>time[i]-0.003 & RawData$NormalisedTime<time[i]+0.003),],paste("Subject.",order,sep=""),summarise,
    #                                MeanX = mean(X,na.rm=T),MeanY = mean(Y,na.rm=T),Time=NormalisedTime[1])

    # As per Cristina's suggestion; fixations are probably more important for pickup of information during eye tracking
    MeanSubjectsTrialFixation[[i]] <- ddply(RawData[which(RawData$NormalisedTime>time[i]-0.006 & RawData$NormalisedTime<time[i]+0.006),],c(paste("Subject.",order,sep=""),"Trial"),summarise,
                                            MeanX = mean(X[Event=="Fixation"],na.rm=T),MeanY = mean(Y[Event=="Fixation"],na.rm=T),Time=NormalisedTime[1])

  }
  MeanResults <- list(MeanSubjectsTrialHit=MeanSubjectsTrialHit,MeanSubjectsTrialFixation=MeanSubjectsTrialFixation)
  return(MeanResults)


}