Even More About Mortality Statistics

I think that it is worth having another go at the statistics – once last time for now. The reason is, a) because there is little to talk about at the moment, and, b) that it is worth considering whether or not the stats which I described are important.

Let me take an example to illustrate what I mean.

It is a fact that, after WW2, there was a gradual but substantial increase in the import of oranges. At the same time, there was a gradual but substantial increase in the number of divorces. The patterns were a near perfect fit. Did the increase in the import of oranges cause the increase in divorces, or was it the other way round? Or were they coincidences? I feel sure that few people would consider that correlation to be other than coincidence. HOWEVER, it is possible that it was not coincidence, and that there was indeed a relationship between imports of oranges and divorces. the relationship could be indirect. For example, it way well be that increasing wealth might fuel both the increase in orange imports and divorces. FURTHER RESEARCH IS REQUIRED! Such research might take place in countries other than the UK, and might also consider the import of bananas and coconuts, to see if there was any connection.

But suppose that there had never been any correlation between the import of oranges and divorces? Suppose that divorces had increased despite no increase in the import of oranges? In its simple way, that would be a good example of the NULL hypothesis, meaning that the starting point of a study would be “That the increase in divorces has nothing to do with the import of oranges”. The study would try to disprove that NULL hypothesis. And it would, since there would have been no increase in the import of oranges.

Now, which would you be more inclined to accept as true, given that there was no increase in the import of oranges, but there had been an increase in divorces? Would you accept the NULL hypothesis or determine to stick with the idea that the import of oranges CAUSED the increase in divorces (or vice versa)?

No one of sound mind would continue to accept that the import of oranges is connected to divorces, but you cannot totally rule it out. Connected, of course, to the idea that you cannot prove a negative.

—-

The significance of the above is that studies which produce evidence that a NULL hypothesis is correct far outweigh any studies which deign to show a correlation. For example, the Enstrom and Kabat study showed that people who were non-smokers and lived with smokers did not suffer more LCs than non-smokers who lived with non-smokers. The WHO study conducted by Boffetta did not show a correlation in similar circumstances concerning heart problems. Both of those studies were huge and used ‘correct’ methodologies. Both found NULL, or insignificant (meaning not worth bothering about), correlations. Both of those studies were ignored by Tobacco Control.

The point is that big studies which produce a NULL result are far more important that studies which produce a result either for or against. The reason is that it is far more difficult for a NULL result to appear when compared with plus or minus in various degrees.

I’m finding it a bit difficult to explain what I mean. Suffice to say this:

I am sure that it is possible to show that a golf ball, striking the head of person, can kill that person. But, in real life, how often do such deaths occur? Should golf be banned because such deaths might occur?

===

I have a reason for this blathering.

A commenter on my last post legitimately produced a calculation of LC death incidence per individual in the UK population, which showed that LC death incidence had reduced, over the whole population of the UK over the last 30 years or so. Well, OK, but…. So what? It is lung cancer DEATHS which are important, and the incidence is not a function of the size of the population, which might well be increased by the import of lots of young people. Diagnosis and treatment also affect deaths.

In the figures that I quoted yesterday, the only variant of any importance is the male/female distribution. I could do the calculations, but what is the point? Even female smoking smoking has fallen from around 50% in 1970 to around 16% in 2000. Or so they say.

===

You get an impression that tobacco control measures are precise because they say so, such as the expectations of PP. The reality is that tobacco control is a blunt instrument. It is a bludgeon.

Why have tobacco companies acceded so readily in Australia? I think that they are part of the experiment.

 

 

Advertisements

2 Responses to “Even More About Mortality Statistics”

  1. michaeljmcfadden Says:

    Peeling oranges is more difficult and less pleasant for women than peeling bananas. After WW2, more husbands began demanding that their wives peel oranges for them. Marriages deteriorated, divorces resulted.

    All because of oranges. It’s perfectly clear to all except the fruit-denialists!

    – MJM

    • junican Says:

      I used a lot of words without really explaining my point properly. Suppose that I put it this way.
      In the last few days, an awful plane crash occurred in the Alps. Everyone was killed.
      Despite every effort that has been made, aircraft crash from time to time. Statisticians can work out the probabilities and would come up with a definite answer to the likelihood of dying in a plane crash. For people who never, ever fly, the Relative Risk of dying in a plane crash is 1 = no risk at all. Anyone who flies at all MUST have a Relative Risk greater than 1. There is no possibility of a negative RR.
      On the other hand, studies of SHS affects have hovered mostly around 1 (1 = non-smokers not exposed to SHS at all). ‘Hovering around 1’ means that studies are showing no effect, and are much more to be believed than propaganda from ASH ET AL.

Comments are closed.


%d bloggers like this: