A method of identification, estimation, and diagnostic checking of autoregressive integrated moving average (ARIMA) models. First, sample autocorrelation coefficients and partial autocorrelation coefficients are used to specify a tentative ARIMA model. Next, the parameters of the tentative model are estimated. Finally, diagnostics tests are performed, e.g. by analysing the residuals. If the model is rejected a new identification is specified and the process is repeated.