On Thu, 21 Jan 2016 12:24:41 +0000, chris wrote:

On 21/01/2016 01:06, Ashton Crusher wrote:
On Mon, 18 Jan 2016 11:44:05 -0800, Jack Black
wrote:

Finally, after years of looking, they found proof that texting causes
accidents!

Here is the quote!

Overall, the hospitalization rate in those states declined by 7 percent
versus states with no bans, the researchers report in the American Journal
of Public Health.

http://www.cbsnews.com/news/texting-...-a-difference/

how do they know texting laws, which I assume are almost completely
ignored anyway, are the cause of the decline in hospitalizations?

Showing cause and effect is really hard with real people in the real world.

In this study they took hospitalisation data caused by car accidents and
tried to measure if there is any difference between states that have a
ban with those that don't. By creating a mathematical model against 8
years worth of data they could test whether the presence of the ban
results in a measurable and meaningful difference. They found a 7%
difference which given the amount of data they had was strong enough to
be explained by the presence or not a texting ban.


Did they do this comparison of the different states for several years
before the ban as well as for a several year period after the ban? If
not I don't see how they can claim their result has any statistically
valid meaning. What were the trends before the ban (that's why you
need several years of data) as well as the trend after the ban. A lot
of studies that claim to show gun control working actually are just
ongoing decreases that were going on before the so called gun control
law and that trend continued after the gun control law. It's a
common statistical error (often done on purpose I think) to get a
politically desirable result.