[alancalverd (OP)]

An advert for yellow spectacles just appeared amongst the good stuff on this forum, with the shock horror headline "40% of driving accidents happen at night!" So obviously I won't buy the specs or drive during daytime.

[jeffreyH]

Well google of all people should see the redundancy of that advert. What with their self-driving cars. Maybe the vision system on their vehicles is faulty. Perhaps?

[chris]

An advert for yellow spectacles just appeared amongst the good stuff on this forum, with the shock horror headline "40% of driving accidents happen at night!" So obviously I won't buy the specs or drive during daytime.

@alancalverd You are priceless. The world would be a much less funny place without you around...!

[wolfekeeper]

In not quite unrelated news apparently 40% of absences occur on either a monday or a friday; which proves that a lot of people are bunking off for the weekends... except that 2 days out of 5 working days is 40% of the days.

Likewise about 50% of the day is, on average night time. So you should obviously do all your driving at night, when it's safer!

[Bored chemist]

It is vaguely plausible that 40% of accidents happen at night, even though most driving is done during the day.

However, unless the glasses stop you being tired at night...

[jeffreyH]

Of course you know what it is. A subliminal advert for Google glass and another for driverless cars.

[Janus]

In not quite unrelated news apparently 40% of absences occur on either a monday or a friday; which proves that a lot of people are bunking off for the weekends... except that 2 days out of 5 working days is 40% of the days.

Likewise about 50% of the day is, on average night time. So you should obviously do all your driving at night, when it's safer!

While it is true that, there are, On average,  an equal number of night time hours and daytime hrs,  It is not true that their are an equal number of  night time "Driver hrs" as day time.   You would have to make a comparison of the total number time people spend driving at night vs. daytime to make sense of the statistic.  So, for example, if people on average do 65% of their driving during daylight hours, then for for the odds of an accident occurring at night to be the same as during daytime, then you would expect 35% of accidents to occur at night, and 40% would be a statistical shift towards higher risk at night.

It reminds me of a poster I saw which tried to make the argument for having sports teams for schools.
It listed a number of extra-curricular activities: Drama, Band, student government, etc along with school athletics and stated that, of all of these, the highest percentage of company CEO's said that they participated in athletics.
This fails on two points:   
Conflating correlation with causation. 
 Participating in sports did not necessarily give these people what they needed to succeed in business.  It could just be that the same characteristics that led to their playing sports also translated into business success.  They might well have done just as well if the sport option had not been available to them.

Improper statistical weighting.
 It did not take into account the relative portion of participation by the general student body in these activities.   For example, you are likely to have a smaller fraction of students participating in student government than you will playing sports. If only 5% of students take a role in student government but 15% play sports, then 3 times as many CEO's would have to say they played sports vs. student gov. just for the odds be equal.

[Colin2B]

This suggests that 49% of fatalities occur at night (USA 2007) with 25% of travel occuring at night. https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/810637

Interesting to look at fatalities with blood alcohol concentration >0.08 g/dL which rises from 14% daytime to 54% nighttime.

This is also interesting as it looks at tests to determine whether information processing for low luminance, low contrast targets is much slower than for well lit high contrast targets. Seems obvious, but good to see it tested.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2564438/