In time series VAR method for multivariate time series analysis OR in SARIMA, is there any way to obtain the forecast on sample data itself i.e train data so that we can plot both forecast and sample data to check how the foretasted results are behaving?
I checked but in most cases, sample data is divided in train and test data set and therefore, I have to do it manually after getting VAR equation and multiplying coefficients with respective data points to obtain forecast results on train data.
Did not get it. Could you please explain a bit. I am doing it in R using VAR package.
Hi Nitesh Jindal
You ca use fitted() command to obtain the in-sample forecasts in vars package.
Example:
mymod
Thanks for your response. I shall try this.I have been using forecast() and predict() but those all were yielding foretasted values. I will try fitted(model)
Thanks
That's right !! The commands, forecast() and predict(), compute the out-of-sample forecast values whilst fitted() produces fitted values or in-sample forecast values.
See chapter 32 of the http://gretl.sourceforge.net/ manual. This explains how to use the package gretl to produce out of sample uni-variate and multivariate forecasts and provides a number of forecast evaluation statistics. You might also find jmulti of use (http://www.jmulti.de/download/help/var.pdf) of use. Both of these packages are free. Similar facilities are also available in many commercial packages.
EViews provides easy-to-use object-oriented feature for in-sample forecast.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training