1
$\begingroup$

How can I in R fit a time series, $x_t$, with external regressors, $v_t$, and an autoregressive error? This time series model is given as follows, $x_t = \beta v_t + \epsilon_t$ where $\epsilon_t = w_t + \sum_{i = 1}^p \gamma_i\epsilon_{t - i}$ and $w_t \sim N(0, \sigma^2)$.

$\endgroup$
3
$\begingroup$

Use the arima() function in the stats package:

fit <- arima(x, xreg=v, order=c(p,0,0))

If you want the autoregressive order selected automatically, use auto.arima() from the forecast package:

fit <- auto.arima(x, xreg=v, seasonal=FALSE, max.q=0)

If you are willing to have more general correlated error structures:

fit <- auto.arima(x, xreg=v)
$\endgroup$
2
  • $\begingroup$ I used this package but I was under the assumption that arima will fit an autocorrelated time series of the form, $$x_t = \beta v_t + \sum_{i = 1}^p \gamma_i x_{t-i}} + \epsilon_t$$? $\endgroup$ – Stereo Oct 14 '14 at 12:56
  • $\begingroup$ Read the help file. $\endgroup$ – Rob Hyndman Oct 15 '14 at 3:09
1
$\begingroup$

gls function in nlme package can do it, with gls(y~x, data=Data.Frame, correlation=corARMA(p=1,q=0))

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.