matlab条件异方差模型,matlab初学-时间序列模型的误差分析（求高手指点） – MATLAB中文论坛...

panlin，你好！我看了你在论坛的其它帖子，注意到你是一个MATLAB高手。

下面程序里面，我选p、q分别为【1 2】、【2  2】、【2  3】，得出【2 2】时AIC最小，但是损失函数和FPE是在【2 3】时最小。

那么我应该怎么选取p、q呢？现在比较糊涂的是：AIC、LOSS FUCTION、FPE各自的意义是什么？

u = iddata(x);

m = armax(u,[1 2])

Discrete-time IDPOLY model: A(q)y(t) = C(q)e(t)

A(q) = 1 + 0.2854 q^-1

C(q) = 1 + 0.2937 q^-1 - 0.0677 q^-2

Estimated using ARMAX from data set u

Loss function 0.000858338 and FPE 0.000894606

Sampling interval: 1

>> aic(m)

-7.0191

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

u = iddata(x);

m = armax(u,[2 2])

Discrete-time IDPOLY model: A(q)y(t) = C(q)e(t)

A(q) = 1 + 0.6283 q^-1 + 0.9642 q^-2

C(q) = 1 + 0.6407 q^-1 + 0.9172 q^-2

Estimated using ARMAX from data set u

Loss function 0.000857283 and FPE 0.000906271

Sampling interval: 1

>> aic(m)

-7.0062

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

u = iddata(x);

m = armax(u,[2 3])

Discrete-time IDPOLY model: A(q)y(t) = C(q)e(t)

A(q) = 1 - 0.011 q^-1 + 0.8905 q^-2

C(q) = 1 + 0.04053 q^-1 + 0.9575 q^-2 + 0.1714 q^-3

Estimated using ARMAX from data set u

Loss function 0.000821395 and FPE 0.000880916

Sampling interval: 1

>> aic(m)

-7.0346