FitRegressionFGN {FGN} | R Documentation |
Fits a multiple linear regression with FGN errors
FitRegressionFGN(X, y)
X |
design matrix, must include column of 1's if constant term is present |
y |
the response variable, a time series |
An iterative algorithm is used to compute the exact MLE.
a list with 3 elements:
Loglikelihood |
value of the maximized loglikelihood |
H |
MLE for H |
alpha |
MLE for regression coefficients corresponding to colums of X |
It is assumed that X is not collinear.
A.I. McLeod
McLeod, A.I., Yu, Hao, Krougly, Zinovi L. (2007). Algorithms for Linear Time Series Analysis, Journal of Statistical Software.
#simulate FGN with mean zero and H=0.2 and fit exact mle for H and mean H<-0.2 z<-SimulateFGN(512, H) mean(z) X<-matrix(rep(1,length(z)), ncol=1) ans<-FitRegressionFGN(X,z) ans #fit a step intervention model to the Nile annual riverflow data data(NileFlowCMS) n<-length(NileFlowCMS) X<-matrix(c(rep(1,n),rep(0,32),rep(1,n-32)),ncol=2) ans<-FitRegressionFGN(X,NileFlowCMS) ans