new ARIMA-model with old ARIMA order

[Teun]

Hi guys,

I have an ARIMA-model that looks like this

txt<-		auto.arima(training, xreg=reg1training, lambda="auto", stepwise=FALSE)
txt
#> Series: training 
#> Regression with ARIMA(0,1,1)(0,1,1)[12] errors 
#> Box Cox transformation: lambda= -0.1176994 
#> 
#> Coefficients:
#>           ma1     sma1    xreg
#>       -0.6507  -0.4964  0.1057
#> s.e.   0.0897   0.1514  0.0188

#> sigma^2 estimated as 0.09377:  log likelihood=-15.75
#> AIC=39.49   AICc=40.16   BIC=48.19

But what if I want to make a new ARIMA-model from the same order.

I know the function

arimaorder(txt)
#>        p         d         q         P         D         Q Frequency 
#>        0         1         1         0         1         1        12 

and

txt$arma
#>[1]  0  1  0  1 12  1  1

Making the new model manually is therefore not difficult.
But how could I accomplish it automatically?

So the end result must be something like

arima(newdata, order=c(same as txt))

Thanks!

[robjhyndman]

Here's an example.

library(forecast)

fit <- auto.arima(USAccDeaths)
fit
#> Series: USAccDeaths 
#> ARIMA(0,1,1)(0,1,1)[12] 
#> 
#> Coefficients:
#>           ma1     sma1
#>       -0.4303  -0.5528
#> s.e.   0.1228   0.1784
#> 
#> sigma^2 estimated as 102860:  log likelihood=-425.44
#> AIC=856.88   AICc=857.32   BIC=863.11

# Use same order, but new parameters
Arima(AirPassengers, order=arimaorder(fit)[1:3], seasonal=arimaorder(fit)[4:6])
#> Series: AirPassengers 
#> ARIMA(0,1,1)(0,1,1)[12] 
#> 
#> Coefficients:
#>           ma1     sma1
#>       -0.3087  -0.1074
#> s.e.   0.0890   0.0828
#> 
#> sigma^2 estimated as 137.5:  log likelihood=-507.5
#> AIC=1021   AICc=1021.19   BIC=1029.63

# Use same order and same parameters
Arima(AirPassengers, model=fit)
#> Series: AirPassengers 
#> ARIMA(0,1,1)(0,1,1)[12] 
#> 
#> Coefficients:
#>           ma1     sma1
#>       -0.4303  -0.5528
#> s.e.   0.0000   0.0000
#> 
#> sigma^2 estimated as 102860:  log likelihood=-531.12
#> AIC=1064.24   AICc=1064.28   BIC=1067.12

^{Created on 2021-03-05 by the reprex package (v1.0.0)}