selection.lag.lead: The lag-lead model selection for MARX function

In MARX: Simulation, Estimation, Model Selection and Forecasting for MARX Models

Description

This function allows you to determine the MARX model (for p = r + s) that maximizes the t-log-likelihood.

Usage

selection.lag.lead(y, x, p_pseudo)

Arguments

y	Data vector of time series observations.
x	Matrix of data (every column represents one time series). Specify NULL or "not" if not wanted.
p_pseudo	Number of autoregressive terms to be included in the pseudo-causal model.

Value

p.C	The number of lags selected.
p.NC	The number of leads selected.
loglikelihood	The value of the loglikelihood for all models with p = r + s.

Author(s)

Sean Telg

Examples

data <- sim.marx(c('t',3,1), c('t',3,1),100,0.5,0.4,0.3)
selection.lag.lead(data$y,data$x,2)

Example output

$p.C
[1] 2

$p.NC
[1] 0

$loglikelihood
[1] -176.9058 -178.0704 -181.4364

Warning messages:
1: In sqrt(df1 * pi * sig1^2) : NaNs produced
2: In log(1 + (E/sig1)^2/df1) : NaNs produced
3: In sqrt(df1 * pi * sig1^2) : NaNs produced
4: In log(1 + (E/sig1)^2/df1) : NaNs produced
5: In sqrt(df1 * pi * sig1^2) : NaNs produced
6: In sqrt(df1 * pi * sig1^2) : NaNs produced
7: In log(1 + (E/sig1)^2/df1) : NaNs produced
8: In sqrt(df1 * pi * sig1^2) : NaNs produced
9: In log(1 + (E/sig1)^2/df1) : NaNs produced

MARX documentation built on May 2, 2019, 3:42 a.m.