bertf 发表于 2013-1-5 21:51 
高人,我又有问题请教了,我按照你说的改了之后,输入了source("lstar.R"),这时候lstar函数就是按照修改过 ...
package tsDyn的架构是class nlar, subclass lstar
所以如果你是要在此架构下修改为estar,
自然需要再source()许多相关的配套函数,
执行过程中,R会告诉你缺哪些函数,你就逐次补足
当然练功是要花时间的,补足了函数,就能运行.
####
source("estar_2013.R")
svpdx=read.table("svpdx.dat")
mod.estar <- estar(svpdx[,1], m=2, d=1, control=list(maxit=3000))
summary(mod.estar)
Non linear autoregressive model
ESTAR model
Coefficients:
Low regime:
const1 phi1.1 phi1.2
0.03546773 -0.02258017 -0.26793628
High regime:
const2 phi2.1 phi2.2
-0.09690616 0.12737871 0.36340519
Smoothing parameter: gamma = 3.671
Threshold
Variable: Z(t) = + (1) X(t) + (0) X(t-1)
Value: -1.176
Residuals:
Min 1Q Median 3Q Max
-3.2464684 -0.3925689 -0.0041534 0.3903538 4.3073348
Fit:
residuals variance = 0.4921, AIC = -654, MAPE = 121.8%
Coefficient(s):
Estimate Std. Error t value Pr(>|z|)
const1 0.035468 0.602035 0.0589 0.953021
phi1.1 -0.022580 0.539471 -0.0419 0.966613
phi1.2 -0.267936 0.124214 -2.1571 0.031001 *
const2 -0.096906 0.614211 -0.1578 0.874635
phi2.1 0.127379 0.531843 0.2395 0.810714
phi2.2 0.363405 0.132885 2.7347 0.006243 **
gamma 3.671356 1.848957 1.9856 0.047074 *
th -1.175598 0.136345 -8.6222 < 2.2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Non-linearity test of full-order ESTAR model against full-order AR model
F = 1.4697 ; p-value = 0.23052
Threshold
Variable: Z(t) = + (1) X(t) + (0) X(t-1)