Unformatted text preview: st.
So you can only get better by combining products. 11/17/12 10 What happens to forecast error if we have a
universal power supply?
The variability of the forecast errors would combine as follows:
σ new = σ 12 + σ 2 + 2 ρσ 1σ 2 where:
σ1,2 are the individual product forecast error standard deviations
ρ is the correlation coefficient of the two errors At start-up the individual errors are
80%+. If the errors tend to offset
each other, then the combined error
will be closer to 40%. Let’s say your individual forecast value is 1.0 and σ1,2 both equal 0.4 (40%). If the errors are completely
uncorrelated (ρ =0), then the standard deviation of the forecast error of the combined product stream would
be 0.57 or 28% of the combined forecast of 2.0
If people tend to buy one product or the other, and a drop in one always occurs with a rise in the other
(perfect negative correlation, ρ = -1), then the new standard deviation would be 0.0
If, say, world events always cause similar errors in both products (perfect positive correlation, ρ = 1) then
the new standard deviation would be 0.80 or 40% o...
View Full Document
This document was uploaded on 03/18/2014 for the course EECS 6.930 at MIT.
- Fall '12