Possible forecast issue about forecast HOT 4 CLOSED

Thanks for letting me know. I'll fix this in the next version. It doesn't
occur for n<=3 because of a different initialization method used when there
are 3 or fewer observations.

Rob J Hyndman
Professor of Statistics, Monash University

www.robjhyndman.com

On Sat, Sep 21, 2013 at 8:32 AM, havik23 [email protected] wrote:

When attempting to run the main forecast function on a time series object
exceptions are sometimes thrown depending on the size of the vector used to
generate the time series object.

For example:

tsObj <- ts(tsVector[1:6], frequency=12, start=c(2013,4))

where tsVector is just a vector of values. Running this time series object
through the forecast function like so:

pred <- forecast(tsObj, 1)

generates the following error:

Error in lsfit(1:maxn, y.sa[1:maxn]) : 0 responses, but only 2 variables
In addition: Warning message:
In lsfit(1:maxn, y.sa[1:maxn]) : 6 missing values deleted

Vectors of length 1,2,3 or >= 12 do not generate an error.

Also, just wanted to say that this package totally rocks! It has been
instrumental in my work and I greatly appreciate it being made available to
the wider world!

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/32
.

from forecast.

Dr. Hyndman,

Thanks for replying. I was wondering if you might be able to recommend a temporary alternative for generating predictions for time series where number of observations range between 4 and 11 using the forecast package? The data set I have contains monthly observations, and in certain instances, I only have somewhere between 4 and 11 observations. Thoughts?

Also, in the cases where n<=3, are you using the meanf function to generate predictions? I was looking through the code and was curious if this was where the different initialization occurs for n<=3.

Thanks again for your time!

from forecast.

Just ignore the seasonal models as follows:
forecast(tsObj, h=1, model="ZZN")

Yes, meanf is used for very small data sets. See
forecast:::forecast.ts for the details.

Rob J Hyndman

www.robjhyndman.com

On Sun, Sep 22, 2013 at 2:25 AM, havik23 [email protected] wrote:

Dr. Hyndman,

Thanks for replying. I was wondering if you might be able to recommend a
temporary alternative for generating predictions for time series where
number of observations range between 4 and 11 using the forecast package?
The data set I have contains monthly observations, and in certain
instances, I only have somewhere between 4 and 11 observations. Thoughts?

Also, in the cases where n<=3, are you using the meanf function to
generate predictions? I was looking through the code and was curious if
this was where the different initialization occurs for n<=3.

Thanks again for your time!

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/32#issuecomment-24865719
.

from forecast.

Now fixed

from forecast.