17.4 Suppose that a particular homogeneous nonstationary time series y, can be modeled as a stochastic process that is ARIMA(1, 1, 1).

(a) How would you calculate the sample autocorrelation functions for y, and its differences and use them to verify that ARIMA(1, 1, 1) is indeed a proper specification for y,?

(b) Suppose you did not have access to a computer package for nonlinear estimation. How would you use a linear regression to obtain approximate estimates of the parameters in the model? (Explain the steps involved clearly.)