How to forecast time series with the help of other training time series in R? Consider such a task: I have the figures on last year's bookselling of a certain book store, and the information of the first half of this year too. Now I want to predict the figure at the end of this year. I've checked R's forecast package. But it seems to only work on a single time series each time. How can I integrate the information from last year? Thank you.
 A: You could also stay within the single equation framework (as opposed to the vector equation framework) and consider building an autoregressive distributed lag (ARDL) model. With two variables, an ARDL(p,q) is defined as:
\begin{equation}
y_{t} = \delta + \sum_{i=1}^{p} \alpha_{i} y_{t-1} + \sum_{j=0}^{q} \beta_{j} x_{t-j} + u_{t}
\end{equation}
and $u_{t} \sim IID(0,\sigma^{2}) ~ \forall t$.
In case it has been overlooked, if you mean that the data you have is bookselling of a certain book store for last year and bookselling of a certain book store for this year, why not concatenate both years of the data (if they are the same variable) so that you have a single series of data? If that's the case then you can take a univariate approach. (As to why the data would be split into two series of data, one for last year and one for this year, I do not know).
A: You could use a vector autoregression (VAR), which allows you to regress on the lagged values of other variables as well as your dependent variable. Or you could include seasonal dummies to account for the seasonality in the data, if that is what you are looking for.   
