Good morning, I was asking a more statistics oriented question on another board and someone demonstrated auto.arima() from the forecast package on my data. The function returned a (2,1,0) model with drift. However when I used the same function it returns a (1,1,0) model. There were no obvious differences in the code. The only thing passed to it was the data. How might this happen? The (2,1,0) model works better so I would like to be able to reproduce the results. Regards, Michael Howell
Different predictions with forecast::auto.arima()
5 messages · Michael Howell, Jeff Newmiller, Abby Spurdle
This is definitely a statistics question still so not on topic here... as changing the data is exactly the kind of thing that can have this effect.
On May 14, 2019 10:51:20 AM MDT, Michael Howell <mchowell2 at gmail.com> wrote:
Good morning, I was asking a more statistics oriented question on another board and someone demonstrated auto.arima() from the forecast package on my data. The function returned a (2,1,0) model with drift. However when I used the same function it returns a (1,1,0) model. There were no obvious differences in the code. The only thing passed to it was the data. How might this happen? The (2,1,0) model works better so I would like to be able to reproduce the results. Regards, Michael Howell [[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Sent from my phone. Please excuse my brevity.
Oh I see. Nevermind then. On Tue, May 14, 2019 at 1:57 PM Jeff Newmiller <jdnewmil at dcn.davis.ca.us> wrote:
This is definitely a statistics question still so not on topic here... as changing the data is exactly the kind of thing that can have this effect. On May 14, 2019 10:51:20 AM MDT, Michael Howell <mchowell2 at gmail.com> wrote:
Good morning,
I was asking a more statistics oriented question on another board and
someone demonstrated auto.arima() from the forecast package on my data.
The
function returned a (2,1,0) model with drift. However when I used the
same
function it returns a (1,1,0) model. There were no obvious differences
in
the code. The only thing passed to it was the data. How might this
happen?
The (2,1,0) model works better so I would like to be able to reproduce
the
results.
Regards,
Michael Howell
[[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
-- Sent from my phone. Please excuse my brevity.
This is definitely a statistics question still so not on topic here... as
changing the data is exactly the kind of thing that can have this effect. I'm sorry. I disagree. This is a question about reproducible code. So, I don't see why it should be considered off topic.
The function returned a (2,1,0) model with drift. However when I used the same function it returns a (1,1,0) model. There were no obvious
differences
in the code. The only thing passed to it was the data. How might this happen?
Either a different version of R, a difference in the forecast package, a difference in one of it's imports (which there are many), or different input. If you're sure the input is the same, then it must be one of the other reasons. I suggest reading the documentation for the relevant forecast package functions. The auto.arima() function alone has many arguments. Changing their values may produce a more desirable model.
Thank you very much for your advice.
On Tue, May 14, 2019 at 11:01 PM Abby Spurdle <spurdle.a at gmail.com> wrote:
This is definitely a statistics question still so not on topic here...
as changing the data is exactly the kind of thing that can have this effect. I'm sorry. I disagree. This is a question about reproducible code. So, I don't see why it should be considered off topic.
The function returned a (2,1,0) model with drift. However when I used
the
same function it returns a (1,1,0) model. There were no obvious
differences
in the code. The only thing passed to it was the data. How might this happen?
Either a different version of R, a difference in the forecast package, a difference in one of it's imports (which there are many), or different input. If you're sure the input is the same, then it must be one of the other reasons. I suggest reading the documentation for the relevant forecast package functions. The auto.arima() function alone has many arguments. Changing their values may produce a more desirable model.