www.bristolmedchi.co.uk/docs/Book%20Review%20by%20Dr%20Dowling.doc

Regards,
Mike Ellwood, aka,
“Montmorency”

Anyone who is interested in his views (whether pro or anti) should be aware that he is not alone.

For example:

www.thincs.org/

www.second-opinions.co.uk/

and all good “Bad Scientists” should really read

“The Diet Delusion” by Gary Taubes, published in the USA as
“Good Calories, Bad Calories”.

Statins are a wonderful example of tinkering with a symptom and having no really positive effect on the underlying condition(s).

The question of who pays for them is almost irrelevant by comparison, except that in many cases we are all paying for them, and I hate my money being wasted.

Regards,
M.

Quite brilliant (thanks Ben).

I should have stuck with Silas Marner.

In the case of a political survey as mentioned above, you could simply pop to the local shopping centre midafternoon IE after workers are back in work leaving mainly local residents in the place, in a staunchly conservative / labor / whatever area and then pick people at random.

The appearance of randomness is there but you’ve skewed the input.

As with all things statistical all possible input bias have to be either taken into account and allowed for or measures should be taken to screen the bias out in the first place.

Also as BG points out, how you phrase a question has a huge effect on how people will answer it.

I could personally re-write the question relating to abortions in the surgeries in half a dozen different ways which would get 6 different returns.

Incidently i think that minor surgery clinics in GP centres are a good idea and i actually had a minor op in one myself once. Much handier than the day care ward in hospitals which require you to be there the whole day for a 20 minute proceedure.

Suggest separate thread, and/or links to clear evidence of misrepresentation.

Then I just get ignored as a pompous arse but at least I’ve made a point!

As an aside, thanks for the tip re the “How to Lie with Statistics” book – looks very helpful. I’m going to get a copy to put with “Straight and Crooked Thinking” on my shelf, although it sounds like the statistics one may be an easier read (I never thought I’d say that about a statistics book….)

People lie when you ask them who they voted for, people accidentally vote for the wrong candidate, people spoil their papers by accident, exit polls miss postal voters, and so on…

All of which notwithstanding, opinion polls tend to be fairly accurate.

Unfortunately consumers of research (like papers) often obsess about sample SIZE to the detriment of sample quality and response rate.

But if they could mug voters in the street and interrogate them using truth drugs, they still wouldn’t beat the sample size problem. Reliable opinion polls are many times more expensive than unreliable ones.

www.washingtonpost.com/wp-dyn/content/article/2008/01/18/AR2008011802675.html

No, not really – the “probability you’re testing for” is the effect size of the environmental factor on the disease, not the prevalence of the disease in the population (although that has a big impact on the sample size as well). So, there isn’t really a good rule of thumb for the situation you’re describing. However, for a population prevalence of 0.001, the required sample size for most reasonable effect sizes (say and odds ratio of 5 or lower) is going to be at least 3000 and more likely much, much larger, so you’re in the right ball-park…

@Humphrey, to say a sample of “less than 1000 (preferably 1500) isn’t really statistically valid” is a bit silly. It depends what you’re trying to prove, really. A sample size of 100 or even 50 could be perfectly adequate for some situations.

Your other point, about weighting the sample if you’re trying to say things about the whole UK population, is much more important.

@danbeck and tin robot, medico-statistical maestros of the great cholesterol circus

Who is the great British sausage eater to believe? Does he believe the statistical porkies of Malcolm Kendrick or the huffy fraudsters of the multi-billion dollar statin-industrial complex? This is one of the great questions of out time: statistically saving British bacon depends on the answer.