FitRegressionFGN: Regression with FGN Errors

In FGN: Fractional Gaussian Noise and power law decay time series model fitting

Description

Fits a multiple linear regression with FGN errors

Usage

FitRegressionFGN(X, y)

Arguments

X	design matrix, must include column of 1's if constant term is present
y	the response variable, a time series

Details

An iterative algorithm is used to compute the exact MLE.

Value

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

Note

It is assumed that X is not collinear.

Author(s)

A.I. McLeod

References

McLeod, A.I., Yu, Hao, Krougly, Zinovi L. (2007). Algorithms for Linear Time Series Analysis, Journal of Statistical Software.

See Also

FitFGN, lsfit

Examples

#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

Example output

[1] -0.008141383
$loglikelihood
[1] 71.21014

$H
[1] 0.2024867

$coef
             [,1]
[1,] -0.004049815

$loglikelihood
[1] -452.3633

$H
[1] 0.7812826

$coef
          [,1]
[1,] 3273.3553
[2,] -570.1046

FGN documentation built on May 30, 2017, 7:19 a.m.