# Not very familiar with odds ratio, should we use odds ratio here?

We have two groups of vehicles (group A: regular vehicle and group B: vehicle with ADAS system), all vehicles are equipped with data acquisition systems for continuously collecting driving data. Participants drove the vehicles as normally as they could, just like drove their own vehicle. 100 participants in Group A drove 10000 miles totally, 5000 acceleration ( >0.3g) were collected, 100 acceleration of the 5000 > 0.7 g, which are treated as safety-critical acceleration. 20 participants in Group B drove 3000 miles totally, 2000 acceleration > 0.3g were collected, 10 acceleration of the 2000 > 0.7 g. Could we use the odds ratio to determine whether ADAS can improve driving performance (less number of safety critical acceleration), like the table below?

I'm not familiar with your application and so the jump from computing a hypothesis test for the odds ratio to "Could we use the odds ratio to determine whether ADAS can improve driving performance" is I (and I suspect many others) will not be able to make.

However, assuming you feel confident enough to make the inference yourself, the table can be used to analyze the relationship between risk of safety critical acceleration and group. There exist several ways of analyzing 2x2 tables like these, and the odd ratio is not my favourite. Odds are difficult to interpret directly and so I'd opt for risk differences because they are easy to understand.

The risk difference is simply the binomial test of proportions (and if I'm not mistaken, a test using odds ratios and a binomial test of proportions should be asymptotically equivalent). There are lots of examples of how to perform this test on the internet and on this site.