ac_corrected {anomaly} | R Documentation |
Transforms the data X to account for autocorrelation by centring and scaling. It uses the transformation X_{i}^{'} = \frac{X_{i}-μ_{i}}{k_{i}σ_{i}}, were μ_{i} and σ_{i} are robust estimates for the mean and standard deviation of each variate (column), X_{i}, of X. The estimates are calculated using the median and median absolute deviation. The scaling k_{i} = \surd{≤ft( \frac{1+φ_{i}}{1-φ_{i}} \right)}, with φ_{i} a robust estimate for the autocorrelation at lag 1, is used to account for AR(1) structure in the noise.
ac_corrected(X)
X |
A numeric matrix containing the potentially multivariate data to be transformed. Each column corresponds to a component and each row to an observation. |
A numeric matrix of the same dimension as X containing the transformed data.
library(anomaly) # generate some multivariate data set.seed(0) X<-simulate(n=1000,p=4,mu=10,locations=c(200,400,600), duration=100,proportions=c(0.25,0.5,0.75)) # compare the medians of each variate and transformed variate head(apply(X,2,median)) head(apply(ac_corrected(X),2,median)) # compare the variances of each variate and transformed variate head(apply(X,2,var)) head(apply(ac_corrected(X),2,var))