Interactive ARIMA Model Diagnostics

This example is a tool created to ease model diagnostics when building ARIMA time series models. There are three key parameters of interest: AR(p), Differencing (d), and MA(q). When building models that have been constructed using these parameters, plots showing the autocorrelation (ACF) and partial autocorrelation (PACF) functions are very useful for assessing whether the constructed models are a good fit.

This tool assists with the process of selecting an adequate ARIMA because it not only shows what the ACF and PACF plots look like, but also transitions the error lines in each plot to new positions. This allows easy comparison between two models because the human eye is very good at detecting movement. Therefore, we can more easily observe differences between two models when there are transitions present than when comparing static images.

The following video demonstrates the arima example found in the sjpMScThesis package, which can be downloaded at my Master’s Thesis page. To run this example yourself, open R and run the following command: library(sjpMScThesis) ; thesisExample("arima")

This video is best viewed in full-screen.

Download

This video example is available in both MP4 (x264) and WebM formats: