What best test to use?

Question

Wondering of any other users have worked with traffic flow data?

The Problem

An access road was closed due to construction works for a month approx.

The hypothesis being tested is " Was there an increase in traffic flow in the adjacent main road?".

My dataset consists in 5 traffic flow sensors at the main road during the access road closure and for the exact same period one the year after with no road closure. Total count of vehicles per day is measured.

I've looked into paired t- test and sampling method ie shuffling.

Any thoughts?

Thank you in advance

Answer 1

Total traffic per day can be assumed to be Poisson distributed - negative binomial if you want to account for additional variability - so you could build a model using as regressors 1) street 2) year 3) week day 4) even year day if you want. then test for year effect. You could also consider interaction between year and day, but test would come out to be different and less straightforward.

However, you should consider that the simpler result is easier to report, and when the problem needs little evidence (of course traffic is affected!) and is less controversial, using a simple, well known test like Student t's will convince everyone and cause you less trouble. Be advised that this is not statistical good practice though.