There is currently a bug in the arima function. Namely, for arima models with differencing or seasonal differencing, the innovation variance estimator uses the wrong denominator whenever xreg is non-null. This is the case, for example, when fitting an ARIMA(p,1,q) model with a drift term (common in financial applications). I reported the bug (and a fix) at https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=16278 , but my report may have fallen through the cracks due to the timing around the 3.2.0 release.
The bug was introduced in the patch displayed here: https://github.com/wch/r-source/commit/32f633885a903bc422537dc426644f743cc645e0 The fix is very simple: https://github.com/patperry/r-source/commit/c1701c05ad91d5631eef196c2007ad9897b01f85 I’ve posted a script that demonstrates the bug at https://gist.github.com/patperry/90a388b056e09cf6a51b Please let me know if there’s anything I can do to help get this fix incorporated. -- Patrick Perry Assistant Professor Stern School of Business New York University ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
