Friday, May 28, 2010

Brush your teeth, save your life.

With the reports coming in that those who never or rarely brush their teeth being 70% more likely to suffer a heart disease you'd expect me to jump all over it; lovely, lovely statistics. Except the newspapers seem to be reporting it in context.

I mean the first thing I'd jump on is whether this is simply causal - that is people who take care of their teeth tend to eat a healthy diet and exercise more so than those who don't; yet medical histories and such were taken into account and this has been reported by the newspapers.

Even more so the newspapers themselves quote the scientists in their speculation that this may be the case and that more studies need to be done.

Next I'd point out the simple '70% more likely' is of no use unless we know the base figures, but yet again although we don't get those figures we do get most of the newspapers reporting that the base chances of heart disease are fairly low.

Could this be a new era in statistical reporting by the media. Is this the end to 'Twice as likely to die' headlines?

I keep expecting to look up and see pigs flying.


Walkerno5 said...

Tell me about it. Watching BBC breakfast, they actually did it full justice. BBC breakfast. Okay, so they also had a segment talking to the cast of Sex and the City, but they reported a scientific finding with the appropriate caveats, and pointed out more studies were needed. I nearly dropped my bacon sandwich.

Don B said...

BBC Radio 4 "Today" programme also handled it very well as well asking lots of questions about control groups and likewise about diet, alternative reasons and what were the caveats to be taken into account. They also asked about where the study was conducted and how large the sample was.

Orphi said...

Wait, let me get this straight… You complain about the media doing statistics wrong, and now you're complaining about the media doing statistics right? ;-)

FlipC said...

I wouldn't say complain, more like shocking :-)

It might be some universal balance as from Liberal Conspiracy we get the recent State of the Nation report that's happy to point out causal connections as meaningful. Then again they've always done that.

Don B said...

@Orphi I had not intended my comment to be seen in any way as criticism but to reflect additional praise for a different set of BBC journalists and editors who had briefed the front of house interviewers.

@FlipC I'm surprised that you've not got a link the BBC Radio 4's More or Less on "the ways we use numbers, statistics, measurements and quantification in our everyday lives" @

Lots of material here for your blogs. I have had to go back over the Banzhaf power index material several times. That piece has certainly made me think hard about what I want out of any alternative democratic system.

FlipC said...

That's cos it be my blog and I've a quirk about audio-only presentations :-)

In all seriousness I do have trouble with audio-only presentations, I can't focus on them.

On of the difficulty with voting systems, as I think I mentioned last time, is not 'is it fair?' but 'will people do it?'. To that end it's all well and good mathematicians stating we should rank all the candidates a la Eurovision, but will people actually do that? How do we allow for mistakes, can you cross out a number, will you need a new form, how long will it take a voter to complete, would that require more stations or longer opening times?

Don B said...

As you point out it's not always what is fair. I must admit Professor Banzhaf came to the conclusion that there is no electoral system that will satisfy everyone. Indeed he concluded most people would be happiest with a benevolent dictator but on the other hand I cannot imagine that you or I would be happy with that. My dictator would be far too tolerant and liberal for all your readers!!!

FlipC said...

Oh I'd be quite happy in a dictatorship, ya know so long as I was the dictator obviously :-)

As I say the difficulty lies in the system, if voters need to turn up at the stations with calculators and working paper they're not going to do it, even the Eurovision contest doesn't work like that, people vote for their favourite, the scores are tallied and the scores assigned accordingly.

Which brings us to the next problem - what is the purpose of the vote? A lot of the negative comments on some of the systems offered is that it doesn't represent the country, the old, 25% of the vote 7% of the seats. Except the current system isn't designed to and was never meant to be representative of the country.

Everyone votes for the candidate to represent their constituency at the national level. Unless you use a system set up for the country (which I don't want) there's always going to be a discrepancy.

The task is to put a system in place that will reduce it while still retaining the 'my MP, my constituency' aspect of the vote.