请教ar(1)模型

shiyiming

刚看到这个很久的留言，其实可以用AUTOREG就可以解决，试试：
proc autoreg data=test;
model y=x1 x2 x3/nlag=1;
run;

个人认为，象时间序列，自相关，联立方程等经济模型，使用EVIEWS是最有效的，功能集中而且齐全，自相关在EVIEWS中可以很简单的就用CORRELOGRAM选项检验，从低阶到高阶。在EVIEWS中，象你上述的模型可以直接用一般线性回归，只用加上自回归阶数就可以了。(y c x1 x2 x3 ar(1))选择普通线性回归选项就能处理了。

shiyiming

I have not checked into it, but if you like to form the problem as a time series question, you can try to use PROC ARIMA in SAS/ETS.  Both MA (miving average) and AR (Autoregressive) are special case for ARIMA model.

shiyiming

非常感谢 onon 和 xic 老师。搁了这么就 ，你们还不吝赐教。

我曾用 EVIEWS 做 时间序列的问题，很好的。但我当时想 sas也应该可以做时间序列的分析，希望都能学习一点，所以在sas中如何实现这个问题就提了出来。

many thanks to you !

romae

shiyiming

I would like you to check out PROC ARIMA in SAS/ETS, it is supposed to be the procedure to handle ARMA model.  Besides, I would like to bring your attention that the AR or MA model of higher order can be re-formaulted as a multi-dimensional model of order 1, by introducing more than one state variable.  For example, write [x(t),x(t-1),x(t-2)] as z(t).

shiyiming

[quote="xic":81300]Besides, I would like to bring your attention that the AR or MA model of higher order can be re-formaulted as a multi-dimensional model of order 1, by introducing more than one state variable.  For example, write [x(t),x(t-1),x(t-2)] as z(t).[/quote:81300]

i am sorry for my ability.I could not understand what's the mean.
Would you like to make some more explanation?

Many thanks!

romae