5
$\begingroup$

Does there exist model (ideally implemented in R) which take into account multiseasonality for multivariate time series as BATS does in the univariate case?

$\endgroup$
  • $\begingroup$ It's very frustrating to see on this site that people can downvote questions which are well-posed and clear without any explanation. $\endgroup$ – paf Jul 12 '18 at 11:49
  • 1
    $\begingroup$ It is a fine question. Some on this site want only high-level, theoretical questions. Others, myself included, enjoy niche questions like this (which can be used by Google-searchers in the future). $\endgroup$ – ERT Jul 24 '18 at 22:46
3
+50
$\begingroup$

Found a possible answer for you at this link, which could provide you with more information on the specific package you are looking for. According to the author, you may want to look into msts as a package for handling your type of data:

An alternative is to use a msts object (defined in the forecast package) which handles multiple seasonality time series. Then you can specify all the frequencies that might be relevant. It is also flexible enough to handle non-integer frequencies.

As an example, below is an example on how to handle multiseasonality, using taylor electricity dataset (from forcast package) with daily ($24 \times 2$) and weekly ($24 \times 2 \times 7$) cycles:

x <- msts(taylor, seasonal.periods=c(24*2,24*2*7)) 
fit <- tbats(x[1:1000]) 
plot(forecast(fit))

Additionally, in case you may want to use dummy variables as @TomWitten suggests, you can view the information on this page, which deals with forecasting double (or multiple) seasonal multivariate time series, with specific examples.

$\endgroup$
  • $\begingroup$ Thanks for your answer (I upvoted it). In fact, although msts objects can handle multivariate time series, that's not the case for bats and tbats so my problem isn't completely solved (except if using multiple linear regression with dummies as explained in the 2nd link you gave). $\endgroup$ – paf Jul 25 '18 at 14:28
1
$\begingroup$

if you know the seasonality could you not just create a series of dummy variables to capture the effects? In my time series data i had 12 monthly dummies for example

$\endgroup$
  • $\begingroup$ That's true of course but that doesn't really answer my question. Thank you (+1). $\endgroup$ – paf Jul 23 '18 at 14:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.