Calibration Flashcards
1
Q
Anagram
A
WBAAP
2
Q
W =
A
What is Calibration?
- Turning theoretical model into reflection of real outcomes
- Makes model accurate to 7-10% (Wood and Tasker, 2010)
- Uses real world data to match observed and predicted data
3
Q
B =
A
Beta
- Distance - euclidian, road distance, cost (time taken)
- Beta controls importance of distance in decision making - high = distance more important, flows smaller - Calibration - obtain known data, compare with predicted, adjust beta so these match better
- Data - Ideally interaction data, but often may not have all data so has to be made as representative as possible. Sources: ATD, Acxiom Data survey, Loyalty Cards
- Disaggeregation - not poss to do per person but can do it on area. Choice (mode of transport, brand) vs. Need (Rural). Use zone characteristics. Birkin et al., 2002 - car market
4
Q
A (1) =
A
Alpha
Attractiveness of store - varies by brand (Clarke et al., 2012), quality, price, accessibility (Fotheringham, 2012)
Market Share is often good for this
5
Q
A (2) =
A
Assessing fit:
Regression - r2
SMRSE - Standard mean root square error
Sum of squares
6
Q
Wood and Tasker (2010)
A
7-10% revenue
7
Q
Birkin et al., (2002)
A
Car Market example of disaggeregation
8
Q
Fotheringham (2012)
A
Store attractiveness affected by agglomeration with other stores
9
Q
Clarke et al., (2012)
A
Store attractiveness changes by brand and area - e.g. Sainsburys