Question

Question 3c: Based on the ME, MSE and MAPE for the 3 different methods shown in the Rankings worksheet, which method would you recommend. Define a weakness and strength of each method, based on the ME, MSE and MAPE and total forecast of population for the year.

		3 different methods chosen
		Regression	HW Additive	HW Multiplicative
ME		20876	4222	-895	Mean Error, but not MAD
RMSE		119540	102288	78353	MSE
MAPE		3.30%	2.60%	1.91%	MAPE
Rank		3	2	1

Answer 1

For ME, HW Multiplicative < HW Additive < Regression
For RMSE, HW Multiplicative < HW Additive < Regression
For MAPE, HW Multiplicative < HW Additive < Regression

So, the best method is HW Multiplicative, second best is HW Additive, and the worst is Regression

	Strength
	ME	RMSE	MAPE	Overall forecast
Regression	-	Unbiased forecast as the square is being considered	-	-
HW additive	Accurate with respect to regression	Unbiased forecast as the square is being considered	Unbiased forecast as the absolute value is considered and relative error is considered	-
HW multiplicative	Most accurate forecast	Most accurate and unbiased	Most accurate and unbiased	Correct choice of model with respect to the time-series plot
	Weakness
Regression	ME	RMSE	MAPE	Overall forecast
HW additive	Biased forecast	Minimizes the MSE after assuming linearity	-	Causality is never assured in the time-series data
HW multiplicative	Inaccurate with respect to the multiplicative model			Nature of plot does not suggest the use of an additive model.
	-	-	-	Complicated in implementation and maintenance as this may be adaptive in nature