Hi there, I'm doing crash count modelling based on road segment. There are two exposure variables in my data, 1. length of the road segment and 2. AADT. How can I include these two f variables in my modelling? Based on Stata, I can declare these variables as a exposure variable, with coefficient constrained to 1 or include these variables in modelling. What is the best way in dealing with these two variables?
There is a lot published on this subject. Too much to mention, you need to do a lit review. Poisson or Negative Binomial models are the norm, although you could explore new frontiers here. SPSS is a good tool, Mini-Tab, R as well. Fundamentals of Safety Performance Functions were given by Ezra Hower. Highway Safety Manual discuses the basics too. Great examples of current best-practice can be found in recent Accident Analysis & Prevention. TRB's Transportation Record has a lot too.
You are correct, both AADT and length are the typical exposure variables on a segment level.
One possible source of information - http://safety.fhwa.dot.gov/rsdp/downloads/spf_development_guide_final.pdf
Thanks Jiri for the sharing. It is very helpful for beginner like me.
Best Regards,
Rusdi
I agree with Chris and Jiri. There are many variables involved in crash such as road width, shoulder width,distance to tree/pole from edge line,sight distance etc. You may try multi-linear regression modelling with many variables in SPSS at the start and narrow down to 2 key variables.
Yes, this is how it is done. Exposure is one part (it HAS to be in model), other explanatory variables (potential risk factors) are the second part (which MAY/MAY NOT stay in the model, after step-by-step removal of the variables with insignificant influence).
Thanks Sinnappoo and Jiri,
Yes. I will adding other variables and identify which one are significant by stepwise technique.
Often the process is not described in detail (it is assumed that readers know by themselves) - one example where it is described very well is here: http://www.ictct.org/migrated_2014/ictct_document_nr_836_24th%20ICTCT%20-%20Tim%20DeCeunynck.pdf (the study is about intersection, not road sections, but the principle is the same).
yes,we can not eliminating total accidental rate or crash rate.using HCS-2000 software for category of LOS in any terrain...this software give Category Of LOS,approach delay, control delay, intersection delay,incremental delay,freeway system,intersection LOS,etc...using this software we analysis average annual daily traffic volume and required width of road.so eliminate or control on accidental rate...
- Badges
- Science topic
- Similar topics
- Computer Communications (Networks)
- Simulators