Reconsider the linear trend data in Table E$4\mathrm{.}4.$ Take the first differ$-$ ence of this data and plot the time series of first differences. Has differencing removed the trend? Use exponential smoothing on the first $11$ differences. Instead of forecasting the original data, fore$-$ cast the first differences for the remaining data using exponential smoothing and use these forecasts of the first differences to obtain forecasts for the original data.