Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I don't really know what's possible, and would like a pointer in the right direction. Many thanks.

I have measurements of time and position which could be anything from a person walking, a vehicle on a road, or in a car park, or a printer in an office. I need to work out journey times for vehicles between two points. It may be they take a meandering route, or even take days to get from A to B. Or they may be a pedestrian, or an emergency service vehicle.

I want the estimated journey time for a normal vehicle along the main route.

The detections are whenever someone is near enough to the detector, which has a particular radius. Sometimes there are very few detections, which probably means the road is empty and journey time would be good, although it could indicate the road is closed, and the journey time would be terrible. Or there could be lots of detections showing traffic not moving, and it could be queueing to turn off the road, but other vehicles are travelling at normal speeds.

The plots look like random noise.


At the moment I am looking at two methods:

  1. Use the interquartile range to discard outliers
  2. Use a Kalman filter.

I think the filter is the wrong way to go as I don't have a model for the journey time other than from moment to moment I don't expect it to change much.

share|improve this question
This looks like a piece of work that requires a methodological paper in Annals of Applied Statistics and a substantive paper in Journal of Transportation Research. You should not expect the community to write both papers for you in neither the answers nor comments, and would rather want to seek the collaboration of a transportation statistician or economist. Or throw this at a grad student in statistics or economics as a dissertation topic. – StasK Nov 6 '12 at 14:52
@StasK I wasn't expecting the community to write papers for me, I asked for pointers in the right direction. Thank you for letting me know that this is a substantial undertaking, though. – Peter Wood Nov 7 '12 at 0:52
I think it is, that's all I wanted to say. If you have uneven time intervals, you may want to consider using variogram modeling and kriging, which generally are considered to be a spatial statistics tools. – StasK Nov 7 '12 at 18:55
@StasK I don't think kriging and variograms are what I need. The geometry of the route between two points is relatively unknown, and unimportant in the kind of results we're looking for. We have two points with detections at each point, and a lot of bogus journey detections. We want to filter out the noise and get a good estimate of current and historical road conditions. Many thanks for your interest. – Peter Wood Nov 19 '12 at 8:48
up vote 1 down vote accepted

I don't know if can give you an expected answer, but I think some Bayesian approach would be good in this case.

You may want to take a look in particle filters instead of Kalman as I susspect it may be a problem to set up correct model for Kalman filter in this case. If you want to go for Kalman, there are different types of the filter and some of them require good knowledge about error covariance, which may cause troubles, but some can compute it with Mante Carlo. Take a look on unscented Kalman Filter.

You may also like as it explain some basic about estimation for moving vehicle and google's self-driving car. (and it is in python).

Maybe some more details in your question would be more helpful and you can get more precise answers.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.