The data plotted below are the (adjusted) amounts (in $$10^6$ ) saved in personal checking accounts in Canadian banks for months between 1976 and 2004 inclusive.
Part (a) Explain why the above series would present difficulties if attempting ARIMA modelling fitting and Box–Jenkins forecasting.


Part (b) Below are plotted the firrst differences of the PCA series above. Which nonstationary feature(s) does the above series exhibit? What would you do to try to reduce this series to a stationary series?


Part (c) Below are plotted the sample autocorrelation and partial autocorrelation functions of a modified version of the PCA series. Comments on the above plots


Part (d) Based on the plots in (c), with the objective of forecasting, would you be comfortable fitting an ARMA model to the adjusted PCA series? Explain your answer.