Another pathological test. arima does not crash for that series that crashes arma: arima(c(2.01, 2.22, 2.09, 2.17, 2.42), order=c(1,0,0)) However, arima crashes for this: arima(c(1.71, 1.78, 1.95, 1.59, 2.13), order=c(1,0,0)) arima seems pretty consistent in its crashing behaviour, since crashing for one series means crashing for all affine series: lets.crash.arima <- c(71, 78, 95, 59, 113) arima(100 + lets.crash.arima, order = c(1,0,0)) # crashes arima(-100 + lets.crash.arima, order = c(1,0,0)) # crashes arima(1 - 0.5 * lets.crash.arima, order = c(1,0,0)) # crashes Curiously, arma does not crash for this second series: arma(c(1.71, 1.78, 1.95, 1.59, 2.13), order=c(1,0)) OTOH, I find it amazing that arma crashes or does not crash for series that differ only on a constant: lets.crash.arma <- c(1, 22, 9, 17, 42) arma(lets.crash.arma, order=c(1,0)) # does not crash arma(20 + lets.crash.arma, order=c(1,0)) # does not crash arma(30 + lets.crash.arma, order=c(1,0)) # crashes Alberto 'Darth Albmont' Monteiro
On Thu, Oct 22, 2009 at 6:19 PM, Alberto Monteiro <albmont at centroin.com.br> wrote:> Another pathological test. > > arima does not crash for that series that crashes arma: > > arima(c(2.01, 2.22, 2.09, 2.17, 2.42), order=c(1,0,0)) > > However, arima crashes for this: > > arima(c(1.71, 1.78, 1.95, 1.59, 2.13), order=c(1,0,0)) > > arima seems pretty consistent in its crashing behaviour, since crashing for > one series means crashing for all affine series:I'm not getting what I'd call 'crashes' with your arma or arima examples- I get an error message and a warning: > arma(c(2.01, 2.22, 2.09, 2.17, 2.42), order=c(1,0)) Error in AA %*% t(X) : requires numeric/complex matrix/vector arguments In addition: Warning message: In ar.ols(x, order.max = k, aic = FALSE, demean = FALSE, intercept include.intercept) : model order: 2singularities in the computation of the projection matrixresults are only valid up to model order1 You've not told us what you get, and the phrase 'crash' normally means some kind of memory error that *terminates* a running R session. Are you really crashing R such that it terminates? In which case, what version number/platform etc? Barry