Table 5.2 shows the estimation results. HHM has a significant improvement over the MNL models both in terms of LL and CAIC. It shows that only one threshold, 2005 m2, is used for representing retail floorspace. Directions with the more retail fioorspace than this number are satisfactory on this factor. The length of the pedestrianized section is represented into three states [<110 m, 110—341 m, > 341 m). Maybe 100 m should be considered as the least length for a pedestrianized street to be constructed. The signs of wb and wd are both consistent with those of the MNL models. The positive β° suggests that smaller overall thresholds were preferred so that alternatives can be differentiated more easily.

Table 5.2: Estimation results of the direction choice models (ENR).

MNL normal variables

MNL logged variables

Parameter

Estimate

Parameter

Estimate

β9

3.466e-6*

βq

0.432*

βι

1.287e —3*

βl

0.137*

βb

0.638*

βb

0.509*

βιί

-0.983'

βd

-1.020*

Nc

2268

NC

2268

Np

4

NP

4

LL

-1048

LL

-1056

CAIC

2131

CAIC

2147

HHM

δq

2005 m2

wb

0.787*

(w9)

1.000*

wd

-6.936'

110m

βe

-3.437*

341m

fir

7.652*

7.452*

β°

4.116*

6.111*

Nc

2268

Np

8

LL

-1002

CAIC

2074

a Parameters that are significant/effective. Only these parameters are counted for calculating CAIC.

Figure 5.2: Distribution of preference structures (direction choice).

Nevertheless, Figure 5.2 shows that risk perception is the dominant force controlling the choice of heuristics. It indicates that the distribution of preference structures concentrates around Ф7. From this point, when preference structure becomes smaller, the probability drops, suggesting that pedestrians tend to avoid using extremely low discriminant thresholds, which would make an alternative preferable to another even with trivial factor advantages, although the decision process tends to be quick. When the preference structure becomes larger, the probability also drops due to the fact that fewer alternatives can be differentiated under such high standards and random choices has to be made, which may give the pedestrians the feeling of losing control. The exception is that the probability of Ф24 is high, even though pedestrians make random choices all the time without considering any information, because this strategy is effortless. In general, pedestrians are risk averse and prefer information search in this particular decision problem. Their way of decision-making approximates rational mechanisms.

The full factorial combination implies 24 factor search sequences, 6 for each factor. The figure also shows the probabilities of factors being searched first aggregated from all the 6 sequences starting from each factor. Under all preference structures, the length of pedestrianized section is always the most probable factor to be searched first. The second most probable first-to-search factor is previous direction. This relationship explains the fact that there are many pedestrians who stopped following their current walking direction into the non-pedestrianized section and turned back at the end of the pedestrianized section. The probabilities of searching floorspace and The Bund first are relatively low. Excluding Ф24, the aggregate probabilities of factors being searched first are, l – 41%, d – 26%, q – 12% and b – 9%.

Found a mistake? Please highlight the word and press Shift + Enter