The ABC Floral shop from problem $3$ is considering fitting various forecasting models on the first seven days of demand and using the second seven days as a hold$-$out sample for comparing the prediction accuracy of the models. They have decided to use $=\mathrm{.}25,$ but aren$$t sure what starting value of forecast, F$1,$ to use.

Try values of F$1=160,$ F$2=170,$ and F$3=180$ to determine the best exponential model for the first seven days using the minimum total absolute deviation as the criterion. You may modify the spreadsheet from problem $8$ for the calculations.

Compare the best model from part a to a three$-$period moving average model on the second set of data. Which one has the smallest sum of absolute errors?

What principles does this problem illustrate?