Simulation function to compute power for AR(1) alternative

Share:

Description

Compares the empirical power of unit-root tests using simulation. Various non-normal distributions may be selected.

Usage

1
2
3
GetPower(phi, n, NSIM = 1000, tests = c("DF", "MLEp", "MLEn", "MCT"), 
	noiseDist = c("normal", "t", "stable", "GARCH11"), df = 5, 
	ALPHA = 1.5, BETA = 0, alpha = 0.2, beta = 0.7)

Arguments

phi

AR(1) parameter or phi=1 if null is true

n

length of series

NSIM

Number of simulations

tests

available tests include: DF for Dickey-Fuller, MLEp for exact MLE using pivotal, MLEn - exact MLEn using normalized, MCT using Monte-Carlo test

noiseDist

distribution of innovations: "normal" for Gaussian; "t" for t-distribution; "stable" for stable distribution; "GARCH11" for GARCH

df

df for t-distribution

ALPHA

shape parameter of stable distribution in (0,2]

BETA

skewness parameter of stable in [-1,1]

alpha

GARCH(1,1) first parameter

beta

GARCH(1,1) second parameter

Value

List with the following components:

power

vector with estimated power for selected tests

phi

AR(1) parameter value

NSIM

Number of simulations used

MOE

margin of error for level 0.95 c.i.

Author(s)

A.I. McLeod

See Also

mleur, dftest

Examples

1
2
GetPower(phi=0.8, n=50, NSIM=100, tests=c("DF", "MLEp"))