R help - Jun 2009 - p-values for ARIMA coefficients

Hi,

I'm a beginner using R and I'm modeling a time series with ARIMA.
I'm looking for a way to determine the p-values of the coefficients of my
model.

Does ARIMA function return these values? or is there a way to determine them
easily?

Thanks for your answer
Myriam

Hi Myriam,

I'll take a stab at it, but can't offer elegance in the solution such
as the more experienced R folks might deliver.

I believe that the ARIMA function provides both point estimates and
their standard errors for the coefficients.  You can use these as you
might a mean and standard error in a one-sample t- or z-test with the
null hypothesis being that the coefficient value is 'zero'.  If the
interval estimate of the coefficient doesn't span 'zero', then you
reject the null hypothesis at whatever level of Type 1 error you
chose.  The R set of distribution functions might be useful for this.

Having offered that as a possible interim solution, you'd be well
served to see what other advice more experienced R users might offer.

Regards,

Cliff



On Mon, Jun 22, 2009 at 6:38 PM, <m.gharbi at yahoo.fr>
wrote:
> Hi,
>
> I'm a beginner using R and I'm modeling a time series with ARIMA.
> I'm looking for a way to determine the p-values of the coefficients of
my model.
>
> Does ARIMA function return these values? or is there a way to determine
them easily?
>
> Thanks for your answer
> Myriam
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

On 23/06/2009, at 11:38 AM, m.gharbi at yahoo.fr wrote:
> Hi,
>
> I'm a beginner using R and I'm modeling a time series with ARIMA.
> I'm looking for a way to determine the p-values of the coefficients  
> of my model.
>
> Does ARIMA function return these values? or is there a way to  
> determine them easily?
The latter.  The arima() function (note the *lower case* letters; R  
is case sensitive)
returns all the info needed to calculate the p-values.

Here's a wee function that wraps it all up:

cwp <- function (object){
#
# cwp <--> ``coefficients with p-values''
#
     coef <- coef(object)
     if (length(coef) > 0) {
         mask <- object$mask
         sdev <- sqrt(diag(vcov(object)))
         t.rat <- rep(NA, length(mask))
         t.rat[mask] <- coef[mask]/sdev
         pt <- 2 * pnorm(-abs(t.rat))
         setmp <- rep(NA, length(mask))
         setmp[mask] <- sdev
         sum <- rbind(coef, setmp, t.rat, pt)
         dimnames(sum) <- list(c("coef", "s.e.", "t
ratio", "p-value"),
             names(coef))
         return(sum)
     } else return(NA)
}

Note that the ``mask'' component of the value returned by arima does  
not get
a mention in the help.  The mask is TRUE where the parameter is being  
estimated
and FALSE where the parameter has been fixed to a predetermined value  
(using the
``fixed'' argument of arima().

######################################################################
Attention:\ This e-mail message is privileged and confid...{{dropped:9}}