Chengke,
> 1.The input traffic flow data - I got 308 counting stations on both major
> roads and minor roads covering Kowloon, the number of sample axial lines
> is
> 214. I do think they can cover the range of both traffic flow and spatial
> configuration variables. I was told the traffic flow data was collected by
> sensor belts installed on the ground road. But I am not sure the detail of
> data collecting. Does that really matter? Actually the data I used was got
> from The Annual Traffic Census 2005, published by Hong Kong Transportation
> Department. According to your words, both the hand gathered and automated
> gathered data have some problem, what data should I use for this kind of
> research.
OK - what you would hope is that these data are self consistent - ie. if
they are counting axles not cars then they are all counting axles,
conversely, if an assumption has been made to translate axle counts to
approximate vehicle counts, then the same assumption has been applied in all
instances. It would be worth verifying with the traffic authority that this
is the case - they should have a specification written down somewhere to
explain the dataset. Now with belts across carriageways you can find that
sometimes different lanes are measured separately, or that different flow
directions are measured separately or together. This information should be
flagged along with the dataset, so it is a matter of asking and checking the
base data. I would also suggest going to a few of the counting sites and
looking to see what is going on.
>
> 2.The axial mapping - It's obviously not an easy process to ensure all the
> underpass and over pass are disconnected, especially in Hong Kong. But
> even
> they are not, the correlation would not be so bad, if the spatial
> configuration really matters.
>
No, if configuration matters, this is very important. The most important
step is to look at the scattergram - a low r2 can be created in many ways
and these often show up in the scatter.
>
> 3.Attribution of multiple counts to a single line segment. Some stations
> locate along one axial line, what do you mean by putting these in as
> separate data points in the stats to start with.
>
Create separate rows in the stats table for each counting site. This will
mean duplicating spatial measures for any sites that are at different points
along the same axial line.
Alan
> Your advices are really helpful, I wish I could get more. Thanks
|