Geography Reference
In-Depth Information
Table 2 Samples of missing
durations estimated from
activities
Estimated from activities
Activity
Estimated duration(hr)
Lunch
0.5
Multi-activity including lunch
1
Food, coffee, ice cream, drink
0.25
Petrol, gas, diesel, fuel
0.15
Gold mining
1.5
T
i
1
¼
T
i
þ
T
i
þ
T
i
þ
1
ð
i
¼
1
;
...
;
n
Þ
:
4.1.2 Estimate Missing Durations from Related Attributes
The basis of this cross reference is that while significant numbers of data items are
not filled in for a specific question, a much greater number of respondents will
have filled in answers to that same question, and this establishes a distribution of
outcomes: for instance for the likely duration of a stop at Franz Josef Glacier. If the
amount known is large enough, which it usually is at significant stops, then this
distribution can be used to model a surrogate duration to replace null responses.
If there are not sufficient neighbouring arrival times for interpolating/extrapolating
in this manner a duration can be assigned using a simpler norm from answered
responses. Table
2
presents the allocated durations used that are based on activity
attributes and are referenced from the highest-frequency of reported durations.
This method can provide a basic form of estimate for activity classes, and can
be modelled in a more sophisticated way using multivariate methods. Once applied
this creates the possibility of a set of partly synthetic but complete timelines
suitable for consistency checking and wider analysis.
4.2 Identifying and Validating Inconsistent Temporal Events
A valid and coherent itinerary is one that sustains a continuous time line which is
consistent with possible travel speed and activities. The method used here for
consistency checking is to calculate a referenced arrival time at each stop based on
travel times from GIS constructed travel paths. The variation between the arrival
time and the modelled value can be taken to check the temporal consistency
between the two stops.
As noted, during survey, tourists tended to report their arrival time with
rounded values, in addition some of the missing durations and missing arrival
times at the first stop after overnight stops were estimated from the high frequency
distribution of known values, these values were soon used to process the GIS
