Lesson 12: Measuring Forecast Error

March 14, 2017

Tue, Mar 14

Review:

• Exam 2 Results
• March Madness Challenge
  • Build a regression model
    • dependent (y) variable = FiveThirtyEight.com Team Rating
    • independent (x) variables = BPI, Jeff Sagarin’s “predictor” ratings, Ken Pomeroy’s ratings, Joel Sokol’s LRMC ratings, Sonny Moore’s computer power ratings and any other data sources you deem appropriate.
  • Generate your own set of Team Ratings and use these ratings to rank every team in the tournament.
  • Write a 2-page paper describing your model building efforts.
  • Model output, rankings and 2-page paper due in class on Tue, Mar 28 (after Spring Break)
  • Be prepared to answer questions about your model to demonstrate understanding
  • Earn back up to 50% of points lost on Exam 2.

Presentation:

• Forecast Error
  • MAD – mean absolute deviation
    • MAD = ∑|Actual – Forecast|/(Number of Forecasts)
  • MSE – mean square error
    • MSE = ∑(Actual – Forecast)^2/(Number of Forecasts)
  • MAD easier to interpret than MSE
• Example
  • Actual vs Forecast Selling Prices

  • Actual	Forecast	|Actual-Forecast|	(Actual-Forecast)^2
    55	55	0	0
    27	50	23	527
    217	184	33	1,097
    153	149	4	19
    180	176	4	12
    145	123	22	477
    105	56	49	2,386
    210	225	15	225
    49	130	81	6,547
    83	104	21	461
    251	272	21	428
    145	166	21	460
    		294	12,639
    		24.5	1,053.3
    		MAD	MSE

Activity:

• Calculate MAD for your Exam 2 regression model forecasts vs actual Sell Price.
• Use a separate piece of paper and submit in-class today.

Assignment:

• Calculate MAD and MSE for your Exam 2 regression model forecasts vs actual Sell Price
• Use Google Sheets
• Submit print copy in-class Thu, Mar 16