Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: The 30 years (plus) "War on cancer"

Gary,

Thanks. Some interesting reports in that set - too many to work through to make quick observations.

Except that I note that the figures that relate to years lost on death appear to assume life expectation to age 80. (That may just be a randomly chosen number for the presentation of course.)

Whilst 80 seems to be a reasonable number these days, based on experiences in my own family circle, it may not be effectively transferable from country to country or even region to region. One notes that for many types of cancer (though by no means all) quite a large percentage of sufferers fall into the 80-84 years category. Would they count as zero years lost or even have a 'gain of years' attribute affecting the overall results?

So many ifs and buts with no clearly consistent and applicable way (in my view) to adjust the numbers to cater for age related matters. If one was able to remove cancer from the list of things people apparently die from, what would be the next major epidemic to take its place? Currently one would have to assume heart related 'something'. Can you image the calls for restricting personal choices and intakes if heart problems became the main 'reason for death'?

Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: The 30 years (plus) "War on cancer"

"So many ifs and buts with no clearly consistent and applicable way (in my view) to adjust the numbers to cater for age related matters."
...............

And so many ways of being creative with statistics,not that govt agencies(trying to justify their budjets) and NGO's(trying to push their adjendas)would do any such things.

For instance,by being somewhat loose with median and mean ages for life expectancy and age of disease death, one can prove all sorts of things.

Definitions:
Median: The middle value of a set of values.

Mean: The arithmetic average, computed by adding up a collection of numbers and dividing by their count.

As an example, let's assume the below list is the result of a search and analysis of guideline companies.
Price/Sales Valuation Multiple
(sorted from lowest to highest)
0.45
0.47
0.49
0.49
0.52
0.55
0.55
0.60
0.61 Middle of sorted sample set
0.62
0.70
0.74
0.76
0.80
0.91
5.30
10.40

0.61 Median
1.47 Mean (Average)
the Median is much more representative of the central tendency of a set. Outliers can dramatically impact the mean (average), whereas the median is less affected.

By using the mean age for life expectancy and the median age for disease death and being loose with the term 'average', one could show large differences.

You'd be right but for the fact that with age of death or age at onset of some disease or other we are talking about populations, not samples. Even if they were samples it would be almost legitimate to treat them statistically as populations because the sets are so large that outliers will have a negligible impact. 71 years is within the ball park of reasonable life expectancy whether measured by mean (bog standard average) or median (half of you will be dead by now) and it's close to the maximum as well. People certainly make it to 71 years without difficulty, but people just don't live 710 years, unless you believe in the Bible that is, and even if they did, adding a handful of Old Testament figures to the mix of 6.6 billion and growing isn't going to change the mean measurably.

So, the best way to answer the question "are lung cancer deaths premature" is to compare the MEAN age at death of lung cancer with the mean age at death of anything. Using standard deviations or even just the raw data, (and since there is a practical upper limit to lifespan which will right-skew the data one should also test that the data is close enough to normally distributed to allow this, or choose a non parametric test, or really push the boat out and pick some complicated method people usually only use when the simple ones give them an answer they don't like), one could even perform a statistical test and play the 5% lottery.

But you are right that one should not ever compare the mean of one dataset with the median of another.