R/getOutliers.r

In OHDSI/Castor: Characterization and Analysis of Statistical Time series comprised Of Real-world data

#'@title Get the outliers for a given time series
#'
#'@description The outliers are computed using different methods based on the \code{outlierType} parameter
#'
#'@usage  tsData.out <- getOutliers(tsData)
#'
#'@param tsData  A time series object.
#'@param outlierType  A string that determines the way to compute the outliers.  outlierType = c("iqr","stl","firstdiff").  Default is "iqr".
#'@details 
#'	iqr: Find outliers using the interquartile range of the given time series (boxplot method).
#'  stl: Use stl weights.
#'  firstdiff: Use the points three standard deviations from the mean of the time series of first diffs.
#'
#'@return A vector of indices which represents the location of the outliers within the time series
#'
#'@export

getOutliers <- function(tsData, outlierType = "iqr")
{

	tsObj <- tsData
	
	if (outlierType == "iqr") {
		tsObj.outliers <-  which(tsObj > quantile(tsObj,probs=.75) + 3*IQR(tsObj))
	} else if (outlierType == "stl") {
		tsObj          <- tsData
		tsObj.stl      <- stats::stl(tsObj,s.window="periodic",robust=T)
		tsObj.outliers <- which(tsObj.stl$weights < 1e-8)
	} else if (outlierType == "firstdiff") {
		tsObj.diff    <- diff(tsObj)
		tsObj.diff.mu <- mean(tsObj.diff)
		tsObj.diff.sd <- sd(tsObj.diff)
		tsObj.outliers <- which(tsObj.diff > tsObj.diff.mu + 3*tsObj.diff.sd)
		tsObj.outliers <- c(tsObj.outliers,which(tsObj.diff < tsObj.diff.mu - 3*tsObj.diff.sd))
	} else {
		stop("Invalid outlier type specified. Valid values are \"iqr\" and \"stl\"")
	}
	
	return (tsObj.outliers)
}