Monday, 2 January 2012

Probability does not exist, claims Cambridge professor

On BBC Radio 4's More or Less on 30 December 2011 Professor David Spiegelhalter, Winton Professor of the Public Understanding of Risk at Cambridge University announced, 'Probability does not exist' .

He defines probability in the programme as 'reasonable betting odds that we construct on the basis of our knowledge and understanding'. But he emphasises that we cannot use probability to tell us, for example,  an individual's risk of developing certain illnesses. Probability as we understand it depends on our knowledge of the situation: we can't rely on it without assuming that we have perfect knowledge.

This supports our view that so-called 'science' based on passive smoking is quite unscientific. If we can't tell a person his or her risk of heart disease, we cannot tell either how much it will be elevated by smoking, drinking, being exposed to smoke or whatever the current fad is. An individual's risk of all these things is also affected by their environmental exposure to risk factors for heart disease, their genetic make-up and other issues that cannot be calculated scientifically.

As far as I understand probability it's a mathematical concept, where the parameters for the different probabilities are known, as they are in the problems presented here. When you can't be sure about the parameters, the probability concept doesn't really work. This applies for much of what is said about smoking as Frank Davis discusses here, and the same applies to any single factor such as consumption of alcohol or junk food (even the definition of junk food is problematic). I once heard of a young woman who had all the symptoms of cervical cancer but was refused screening on the grounds that she was unlikely to have it on the grounds of her age. Here is a case where the low probability of breast cancer in men resulted in the denial of treatment to a man who had (100 per cent probability of) breast cancer.

Professor Spiegelhalter will take his ideas further at this event (evening/smart dress recommended).


Anonymous said...

I'd say that there is only a 50% probability the Prof is correct.

It is quite common for unknowns to be allowed for.

That is why you see things like +/- 10% in conclusions.

People also think that probability can assure certainty.

If you roll a dice, there is a 1 in 6 probability of rolling a 6.

But; if you roll 6 dice at the same time, there is NOT a 100% probability that one of them will come up 6.

There is a 33% probability that there will be NO 6's.

I THINK(?) that there is about a 1/7,800 probability of a single 6 coming up.

Anti 'science' is bad because they draw conclusions that are not warranted by their data.

For instance; cancer and heart disease cause about 50% of the yearly deaths, so antis say that 50% of the smokers will die from a disease 'caused' by their smoking.

In the USA there are about 94 million ever-smokers and antis say that smoking 'causes' about 400,000 smokers deaths per year.

47 million(50% of 94 million) divided by 400,000 = 117 years.

Soooo, antis seem to be claiming that smokers live,on average, to be 117 years old.

Gary K.

Anonymous said...

" But he emphasises that we cannot use probability to tell us, for example, an individual's risk of developing certain illnesses."

But, we can.

In America, a never-smoker exposed to SHS/ETS has a yearly 1/20,000 probability of dying from lung cancer.

Thus; we can, tell such an individual that THEIR yearly probability of NOT dying from lung cancer 'caused' by SHS/ETS exposure is 99.995%.

19,999/20,000 = .99995 = 99.995%

We CAN tell such an individual that there is ONLY 5/1,000th of a 1% yearly probability of their dying from such a lung cancer.

Table 4

American CDC data shows us that the average(median) age of death 'caused by' smoking is 72.

Thus, we can tell the above individual that if they are that 1/20,000, there is a 50% probability their death will not occur until they are OVER the age of 72.

Table 4 also tells us that there are NO such smoking 'caused' deaths below the age of 35 and 95% of them occur over the age of 45.

Let's see how many cigarettes smoked it takes to 'cause' a death from smoking, if it is to occur.

A,pack per day,18 year old smoker will smoke 7,300 cigarettes per year.

That 18 year old smoker will have smoked 124,100 cigs before they each the age,35, at which such deaths START to occur.

That 18 year old smoker will smoke 197,100 cigs before they reach the age,45, at which there is a 95% probability they have NOT died from such a death.

By age 72, our smoker will have smoked 394,200 cigs and there is STILL A 50% probability that death has NOT occurred.

Probability can tell us quite a lot!!!!

Gary K.