I am working with the multivariate hourly electricity price data of six years data in which the first five years are use as testing set and the last year as validation set. I want to forecast the last year through functional autoregressive model of model more than one and then check the accuracy of the model . As per my knowledge or if I wrong then please correct me that their no statistical package available in R for functional autoregressive model of order more than one. MY objective is one year forecast (mean24*365) through generalize functional autoregressive model. So, I trickle my problem as first I transform the multivariate data into functional data through Fourier basis function, then calculate the functional principle components as dimensional reduction, choosing the first six functional principle components explaining a given a mount of variation. Then taking the corresponding six functional principle scores and apply vector autoregressive model, choosing the dimensional and the order through functional final prediction error, the forecast function is used for one day a head forecast , Know as i choose the first six functional principle components so i obtained the six forecast values but that how to obtained these 24 forecasted values of a single day and then transform the vector autoregressive into functional form KL transform, how i can generalize my result for one year(24*365), if i am wrong then please suggest me how to trickle my problem. i sharing my code to understand my problem easily. , How i check the accuracy of the model.
You might find this link of interest. http://msekce.karlin.mff.cuni.cz/~vorisek/Seminar/1314z/1314z_Cerny.pdf
Good luck, D. Booth
Let me recommend this framework, it can be used to decompose your problem and to evaluate the accuracy of your model:
Conference Paper Data Formats and Visual Tools for Forecast Evaluation in Cyb...
The key approach here is to use rolling-origin evaluation and to store forecast result in TSTS and FTS schmeas, as described in the above conference presentation.