0
$\begingroup$

I have an assignment, where I'm supposed to measure whether an intervention at a border crossing point will an effect on the traffic flow.

We have a total of 7 border crossing points and we continously measure the traffic flow (n vehicles pr. hour).

Intervention period will be 3 weeks at one border crossing point.

My initial thought was to basically compare means (t-test) between intervention period (3 wks) vs non-intervention period (3 wks up to intervention). The problem is however that traffic flow is stochastic, hence flow changes could easily be contributed natural variation. Is it possible to some how adjust for these natural variations using the remaining 6 border crossing points?

$\endgroup$
0
$\begingroup$

I think the problem in this question raises on a regular basis. Mostly you are not able to isolate your experiment in the way that there are no other effects, than the one you want to investigate.

You are talking about stochastic variation in the traffic flow, so let's take about time scales. Do you see variation over hours, days, weeks. The longer these periods are relative to your intervention period, the more important for further considerations. When I hear variations in traffic flow I first think of holidays or similar. (Naive view: try to avoid times with obvious and distracting actions)

This leads to the next point how to overcome the variation. Also trends could be possible in an other context. As you have the data of all seven border crossing points, you can try to estimate the flow on one points using the others. If the results are promising you can use the results in the intervention period, too. Include also estimation errors in further processing. Nevertheless this way can make problems if your intervention also has side effects to the other crossing points. Like, this extreme example: closing one border crossing will increase the flow on a neighboring border crossing.

The estimated traffic flow without intervention can be compared to the traffic flow with intervention in known ways as you said.

$\endgroup$
1
  • $\begingroup$ First of all, thank you for your reply. Variation may not be that stochastic, as periodic variation is quite obvious on daily basis and yearly basis. If I draw trend lines, both are bell-shapped with peaks mid day, mid year, respectively. Only exception is holidays, where n vehicles are generally lower. $\endgroup$ – Magnus Hoffmann Mar 6 '20 at 10:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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