A Shiny Application as a Teaching Tool on Forecasting Methods

Abstract

In this paper, we develop and propose a Shiny application focusing on teaching forecasting methods. Shiny is a statistical computing framework based on the software R, which is used to build several web applications. The developed application contemplates several forecasting methods namely: classical decomposition (additive and multiplicative); naive forecasting; exponential smoothing (automatic selection); and the autoregressive integrated moving average models (ARIMA), allowing the user to fit different models on data that can be imported from spreadsheets and produce different plots in a friendly, open and accessible interface. Further, we can also evaluate the fitted models through a number of different residuals plots (dispersion, histogram, autocorrelation, and partial autocorrelation functions), and produce forecasts through the application. We can conclude that the developed application, based on the Shiny framework, is an interesting alternative for the improvement of statistical teaching, as well as in applied research when we are dealing with forecasting methods.