Question

R代码：

  library(forecast)
  library(FitAR)
  set.seed(54321)
  c<-0 # Counter for correct model under method
  n=50
  nsim=10
  phi=c(0.5,-0.2)
  t=1:n
  t.square=t^2
  beta1=2
  beta2=3
  beta3=4
  for (i in 1: nsim) {
  error<-arima.sim(model=list(ar=phi),n=n,innov=rnorm(n,0,1))
  yt=beta1+beta2*t+beta3*t.square+error
  dt=cbind(t,t.square)
  p<- SelectModel(as.ts(error), lag.max = 15, Criterion = "BIC", Best=1)
  f1=Arima(yt, xreg=dt, order=c(p,0,0),include.drift=FALSE, include.constant 
  = FALSE, method =  "ML")
  print((coef(f1)))
   }

输出：

       ar1        ar2        ar3          t   t.square 
    0.9031324 -0.4968706  0.4069485  3.1808610  3.9965791 
       ar1         t     t.square 
     0.7372397 3.1051317 3.9989081 
                    t    t.square 
               3.178753 3.996895 
        ar1         t     t.square 
     0.6279603 3.1813097 3.9967204 
        ar1         t     t.square 
     0.5789377 3.1561776 3.9976448 
      ar1        ar2          t     t.square 
     0.8023629 -0.2414305  3.1717066  3.9968250 
      ar1        ar2          t   t.square 
     0.8423128 -0.3565319  3.1768333  3.9966517 
     ar1         t     t.square 
     0.5170698 3.0990545 3.9987464 
     ar1         t        t.square 
     0.5521029 3.1356383 3.9978553 
     ar1         t        t.square 
    0.5280407 3.1679048 3.9972218

模拟数量为10。上面的输出是对2阶AR的估计，变量为t和t.square。

我想让代码在AR（2）有两个系数时得到。这意味着对于每个模拟，如果ar1和ar2存在，则不删除该行。

所以我想得到

      ar1        ar2          t     t.square 
     0.8023629 -0.2414305  3.1717066  3.9968250 
      ar1        ar2          t   t.square 
     0.8423128 -0.3565319  3.1768333  3.9966517

先谢谢您。

Answer 1

我认为这可以解决问题。首先，可以在while上使用i循环，仅在获得所需结果（AR（2））时递增计数器。然后，您可以确定系数矢量的名称包含ar2而不包含ar3的条件。然后，仅在那些情况下才保留模拟值。

library(forecast)
library(FitAR)
set.seed(54321)
c<-0 # Counter for correct model under method
n=50
nsim=10
phi=c(0.5,-0.2)
t=1:n
t.square=t^2
beta1=2
beta2=3
beta3=4
out <- array(dim=c(nsim, 4))
colnames(out) <- c("ar1", "ar2", "t", "t.square")
i <- 1
while(i <= nsim){
    error<-arima.sim(model=list(ar=phi),n=n,innov=rnorm(n,0,1))
    yt=beta1+beta2*t+beta3*t.square+error
    dt=cbind(t,t.square)
    p<- SelectModel(as.ts(error), lag.max = 15, Criterion = "BIC", Best=1)
    f1=Arima(yt, xreg=dt, order=c(p,0,0),include.drift=FALSE, include.constant 
        = FALSE, method =  "ML")
    co <- coef(f1)
    insim <- ("ar2" %in% names(co)) & !("ar3" %in% names(co))
    if(insim){
        out[i, ] <- co
        i <- i+1
    }
}

使用R删除行

1 个答案: