I am using data of 25-year interval, one dependent variable and 5 independent variables, all are time series in nature. I would be grateful for a step by step help in the process of analysis with STATA using the ARIMA model, do I differentiate all the independent variables to test for stationarity or I only do that to the dependent variable? Thank you.
Hi,
Pls see below for the steps
1. Check the data for structural breaks
2. Check stationarity of data (Unit root tests) - if data is not stationary, transform into stationary form (1st difference or so)- consider log transformation as well.
3.Check the seasonality of the data (Seasonal unit root tests or seasonality index)
4. Identify the appropriate AR and MA terms (if seasonality is present, add seasonal AR terms and seasonal MA terms) use information criteria such as AIC, SIC,BIC for selection)
5. Estimate ARIMA or SARIMA model
6. Diagnostic tests( AC, PAC graphs, Q tests, normality tests, etc.)
6.Examine in-sample forecast accuracy
7.Out of sample forecast accuracy
You may also refer the following article by me.
Conference Paper Forecasting International Tourist Arrivals To Sri Lanka Usin...
By the way, when fitting an ARIMA model, there is no dependent variable or independent variable. We take only one series at a time for one model as ARIMA is a kind of univariate modelling.
Kind regards
Thushara
you can follow stata's manual: https://www.stata.com/manuals13/tsarima.pdf
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series