R中如何给时间序列模型做参数检验？

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

建立了模型，做了参数估计后，如何给参数做检验呢？函数怎么写？谢谢达人，下面是我的程序
a<-c(15.48,14.85,15.29, 15.60,17.16,16.10,17.71,19.20,17.28,17.19,16.97,
17.30,16.93,17.93,17.65,18.58,18.03,16.90,15.31,15.65,
14.42,14.70,14.59,15.36,15.32,15.04,15.91,16.24,17.00,
17.25,16.65,15.84,14.26,14.74,14.01,15.30,15.29,15.61,15.82,
15.60,15.17, 15.60,15.05,16.00,15.90,16.56,16.75,17.02,16.55,
16.66,16.37,16.59,16.92, 16.59,17.07,17.68,17.79,17.58,17.42,
17.42,17.48,18.63,18.64,18.88,19.46, 18.76,17.44,17.93,18.44,
17.90,18.36,18.56,17.95,18.08,19.89,21.88,21.55,22.39,24.06,
22.97,15.72,16.53,16.60,16.60,17.40,18.25,18.60,18.32,18.45,
19.47,18.70,17.41,18.20,16.38,17.90,18.75,19.00,18.55,18.61,
18.54,18.55,18.95,19.10,19.97,20.80,
20.06,20.23,19.66,19.28,19.40,20.18,19.40,19.12,18.64,19.41,
19.32,18.56,19.38,19.93,20.52,21.20,21.26,20.56,22.62,23.47,
23.31,25.64, 26.36,27.05,27.25,27.41,27.30,28.49,28.30,28.70,31.57,
33.80,33.97,33.38,32.27,35.34,33.76,33.13,33.30,33.72,32.08,
30.85,33.07,32.93,34.00,34.79,34.00,33.58,33.80,33.70,34.13,
33.08,33.29,31.98,32.40,30.78,30.97,31.93,31.93,30.70,30.80,30.46,
29.90,30.99,31.53,29.86,29.25,30.20,33.22,33.21,33.30,34.27,32.89,
31.27,32.26,31.60,30.60,31.50,31.70,33.00,33.51,33.80,35.45,
35.82,35.67,38.93,39.50,39.05,35.90,36.08,36.50,34.93,35.00,
36.37,35.28,36.37,36.00,34.30,32.93,33.30,32.82,31.18,31.50,
31.46,30.20,29.96,31.26,31.55,32.53,32.92,32.60,33.50,33.52,
33.10,32.30,31.02,29.40,29.95,27.10,27.25,27.80,28.38,28.30,
28.66,27.83,27.74,29.31,28.84)
ts(data=a,start=2007,deltat=1/233)
ts.plot(ts(data=a,start=2007,deltat=1/233))
acf(a)
ts.plot(diff(a,2))
b<-diff(a,2)
Box.test(b,12,type="Ljung-Box")
acf(b)
pacf(b)
m1 <- arima(a, order=c(5,2,1),)
tsdiag(m1)
plot(resid(m1))

round(pnorm(-abs(m1$coef),sd=sqrt(diag(m1$var.coef))),5)  /*参数估计*/
请问如何给参数做检验呢？是用summary（）么？还是怎么做？达人请教下，谢谢。

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：时间序列模型 时间序列 参数检验 参数估计 时间 检验 模型 序列 参数

达人啊 - -

没人理我 5555555555

怎么会没人理我的- -我没论坛币可以报答0 0~太惨

时间序列比较复杂，我就我所了解的列出一下要进行的检验和方法吧。

单位根检验：urca::ur.df(x,lags,type),这是进行一阶单位根检验。用来证明时间序列的平稳性。如果一阶不平稳，就要进行二阶单位根检验，首先进行差分Lx<-diff(x),然后urca::ur.df(Lx,lags,type)。

如果是多元时间序列，最好做一下协整检验，用来发现多个时间变量之间是否存在关系。
最简单的是Pillips-Ouliarisu协整检验：
首先是方差率检验：pu.test=summary(ca.po(tsdataframe,demean=’const’,type=’Pu’))
然后是多元迹检验：pz.test=summary(ca.po(tsdataframe,demean=’const’,type=’Pz’))
其中tsdataframe 是要把你的时间序列变量转为专门的格式window(cbind(x1,x2,x3),start=,end=)
Johansen方法也可以进行协整检验，稍微复杂点，不过它可以告知存在几个协整向量。具体操作自己看书。

如果前面的检验工作完成，就可以进行VARX和状态空间模型的拟合了。