# How to perform Timeseries Forecasting and Modelling For subsets?

Hi I have a problem with my project. I have 1000 data and I used it as training data to forecast with Timeseries. But my data have some subsets and it's impossible to timeseries forecasting and modelling in R . I already searched from the Internet and asked some experts that R doesn't support Timeseries Model and Forecasting for subsets. But one of the requirement for my project, I must use R as statistic tool.

Do you have any suggestion or R script which will support Timeseries model and forecasting for subsets in R?

• you have lots of run-on, and not a lot of clarity. what was the general form of the data. why is it not amenable to use in ts modeling? there is plenty of time-series forecasting in R, I guarantee it. did you look on CRAN? – EngrStudent Dec 8 '15 at 3:57
• @EngrStudent Ya i want to use ts modelling. I thought R doesn't supported subsets. My data's pattern is double seasonal. The other statistic tool that supported modelling subsets, is SAS. Do you have any suggestions? – Shieryn Dec 8 '15 at 4:35
• By "subsets" do you mean (this)[stats.stackexchange.com/questions/6239/…? – EngrStudent Dec 8 '15 at 11:37
• @EngrStudent I mean, what R syntax that needed to TS modelling with subset ARIMA. The usual ARIMA(p,d,q) but when TS subset is like ARIMA ([1][2][8],0,[7][9][10]). how? – Shieryn Dec 8 '15 at 11:49
• @EngrStudent Oh ya like that. I hope my data only have subsets in AR. Idon't know what to do if my data could've subsets in MA. Thankyou for your help :D – Shieryn Dec 10 '15 at 3:53

## 1 Answer

I'm not sure about what you mean, sorry if i'm not in right way. Are you working in Panel Data? is that or just simple subsets you want perform forecasting? Like working with "structural broke" when you crop your data in 2 periods to analyses ?

I had problems to use some packages + ts, so must managed my data like this:

my_ts<- as.numeric(as.character(my_ts\$Value1))  % you can subset here, if is a simple subset (not panel data)


then

my_ts<-  ts(my_ts, start = c(1994, 1), end = c(2014, 1), frequency = 1)

• noted that you're looking for MA. Didn't the auto.arima() give any hint about? – Adilson V Casula Aug 2 '17 at 11:47