[R] Why does ar.ols manual caution against intercept=T, demean=F

[Bauer, Michael]

The manual for ar.ols {stats} cautions against fitting a model with intercept to a time series that is not demeaned. From the manual:

"Some care is needed if intercept is true and demean is false. Only use this is the series are roughly centred on zero. Otherwise the computations may be inaccurate or fail entirely."

I was surprised by this warning. If a series is not centered on zero/not demeaned by ar.ols, I would typically include an intercept in my AR/VAR model. This takes care of fitting the unconditional mean.

Is this a typo? It would make sense to warn against fitting a model with "intercept=F, demean=F" for a series that is not centered around zero.

I frequently fit AR/VAR models with intercept and without demeaning, couldn't find any more information on this aspect of ar.ols, and therefore am asking my first question on R-help (after several years as a very happy R user.)

Thank you,
Michael
-----------------------------------------------------------------------
Michael D. Bauer
Economist, Research Department
Federal Reserve Bank of San Francisco
101 Market Street MS 1130
San Francisco, CA 94105
415.974.3299 (w)   415.471.9136 (c)
michael.bauer at sf.frb.org
http://www.frbsf.org/economic-research/economists/michael-bauer/