I'm far from been an expert on the subject. But is this a bug in TimeSeriesModelFit?
If I execute this code.
data=Range@10;
tsm=TimeSeriesModelFit[data,{"ARIMA",{1,1,1}}]
ListLinePlot[{tsm["TemporalData"],TimeSeriesForecast[tsm,{10}]}]
I get an error: "None of the candidate models could be successfully fitted to the \ data. Try a different model family or set of candidate models"
If I just change data to:
data=Range@10+0.001;
I get the correct answer:
![enter image description here](/c/portal/getImageAttachment?filename=CapturadeTela2014-07-05%C3%A0s00.17.29.png&userId=25532)