Keywords - Function groups - @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Library: times
See also: armalik armapq

Quantlet: armacls
Description: estimates an autoregressive moving average process with zero mean by maximizing the conditional sum of squared residuals.

Reference(s):

Usage: {y,wnv} = armacls(x,p,q{,maxiter,eps})
Input:
x n x 1 vector, the process (time series)
p scalar, the autoregression order
q scalar, the moving average order
maxiter (optional) scalar, maximal number of iterations; default: maxiter = 250
eps (optional) scalar, convergence tolerance of the function value; default: eps = 1e-7
Output:
y list containing 1. (p+q) x 1 vector, the estimated parameters, 2. the number of iterations, and 3. a scalar indicating whether convergence occurred (=1) or not (=0)
wnv scalar, the estimate of the white noise variance

Example:
library("times")                      ; loads the quantlets from Times Library
randomize(0)                          ; sets random seed
x = genarma(0.7|0.1,0.3,normal(500))  ; Generates ARMA(2,1) with White Noise
{y,wnv} = armacls(x,2,1,50)           ; Estimation procedure for above ARMA(2,1) process
y
wnv
randomize(7)
{y,wnv} = armacls(x,2,1,50)           ; Reestimation with different start values
y
wnv

Result:
Contents of y.minimum           ; vector with estimated parameters
[1,]  0.56378                   ; AR(1)
[2,]  0.22206                   ; AR(2)
[3,]  0.40891                   ; MA(1)

Contents of y.iter              ; Number of iterations till convergence
[1,]       32

Contents of y.converged         ; Convergence achieved: YES
[1,]        1

Contents of wnv
[1,]  0.91065

Contents of y.minimum           ; vector with estimated parameters
[1,]  0.56627                   ; AR(1)
[2,]  0.22044                   ; AR(2)
[3,]  0.40804                   ; MA(1)

Contents of y.iter
[1,]       44                  ; Number of iterations till convergence

Contents of y.converged
[1,]        1

Contents of wnv
[1,]  0.91065



Author: C. Hafner, W. Haerdle, 19960906, G. Aydinli, 20020118 license MD*Tech
(C) MD*TECH Method and Data Technologies, 05.02.2006