Geography Reference
In-Depth Information
levels of the stop; about 4 arrival times that are not consistent with prior temporal
events, and 12 % reported arrival times reported earlier which are later than the
estimated times. These notably early and late reported arrival times were tagged as
suspect temporal events in the WCTFS dataset.
Legs with possible missing stops were then inspected in respect of the reported
late arrival times, which possibly are inaccurately late or are accurate but conceal a
missing stop of some length. During the survey, respondents were required to
record any stops including stopovers with duration of 5 min and more: we know
that not all stops get identified at all. Several criteria might alert us to
mis-reporting, including the claimed length of continuous driving and the claimed
duration of the activity at the new stop. Excessive driving time between stops may
provide some clue to the existence of misunreported stops. Chen ( 2004 ) identified
that the typical continuous driving time of tourists is close to 2 h. We reason from
this that sections of continuous driving over 2 h with no stops recorded are more
suspect than shorter sessions in terms of possibly concealing a missing stop and its
duration. Stop to stop travel which exceeds the estimated time by more than
30 min is tagged as highly likely to have failed to report at least one stop and the
stops with such arrival times are interrogated as inconsistent temporal records.
There are a total 159 records out of a total 21,084 that were tagged in this category.
3.2 Patterns in Spatiotemporal Data Quality
We know that the completeness and consistency of the reporting of spatiotemporal
events varies from one person to another during a survey. Variation in reporting
may occur due to different people having different levels of ability to recall an
event. Respondents from different gender groups, age groups, social classes or
countries of origin may exhibit different biases of interest and so contribute to
different levels of reported data quality. The nature of the stops and their associ-
ated events also influence reporting, leaving different impacts on respondents'
memories. Some events are easier to retrieve and therefore reported by respon-
dents accurately. In contrast, some events may make less impression and enjoy a
lower level of recognition with respondents and are likely to miss out or be
recorded incorrectly. In short, many variables affect reporting rates and accuracy.
To use the data within a weighted analysis and look for appreciable patterns or
relationships we typically assume that error is randomly generated. We might also
hope that absence is similarly distributed with spatiotemporal data. It seems
possible and is hoped that error issues and data omissions exhibit regularities that
would be valuable to identify. One example would be the increase of missed or
poor data with time into the itinerary. Time lapse for recall between experience
and reporting is a similar case. This section reviews the revealed patterns of non-
reporting (absence) and apparent anomalies in reported data items.
Search WWH ::




Custom Search