solar prediction using ARMA

Hi every one, I am working on a time series analysis and want a time series data to be modeled using ARMA i have 8 years of time series (with missing data) for six solar stations , what i want to know: how to choose the train and test groupes (how many years to use for each groups, how many stations i have to use to do the simulation and get arma parametres)? Are there criterias? Or I need do my simulations for a period of time on each station (For example the first 4 years of my datas) and looking for the most repeated settings? do i need to fill the missing data? i am really lost here, can any one help me please.