Fitting a reduced-form MA(3) time series model in R

Question

I am trying to fit an ARIMA model for a certain financial time series. I've used EViews for modeling, and have decided to fit a so-called reduced-form MA(3) model, where only the third lag is statistically significant.

Unfortunately, I have not been a to figure out how to do this in R. All I can find is how to fit a regular MA(3) model, using either the 'stats' or 'forecast' packages.

Can anyone please help me out? Thank you!

NOTE: This is a cross-post from SO. The good folks there suggested that I use the argument 'fixed' for the arima{stats} and Arima{forecast} packages, but I can't figure out how to use it just from the documentation.

Answer 1

I believe you'd do something like:

f <- arima (USAccDeaths, order=c(0, 0, 3), fixed=c(0, 0, NA, NA))

which sets the first two MA coefficients to 0, and allows the third MA coefficient and your intercept to vary. Not sure if that will give you the same coefficients you got in EViews or not, but it's what I'd try first.