[R] t-stat for the coefficients of an ARIMA model

[David Winsemius]

On Nov 25, 2010, at 10:50 AM, Samuel Le wrote:

> Dear all,
>
>
>
> I am  fitting a time series using the following command:
>
> Ts.arima<-arima(x,c(2,1,2)) where x is a time series.
>
> What the function returns is perfectly fine but I was wondering if I  
> could access to the t-stat of the coefficients I got from the arima  
> function.

The typical approach is to see if there is a coef function and a vcov  
function for your fit and to see if this gives sensible results:

coef(fit)/sqrt(vcov(diag(fit)))

(I looked at the docs and those functions are available for arima  
objects. However, it's still your responsibility to properly interpret  
such output. I have no substantial experience with time series  
analysis and as a general rule worry when the package authors choose  
not to provide a particular statistic. )

David Winsemius, MD
West Hartford, CT