On Monday, February 19, 2024 at 1:21:55 AM UTC+11, Fred Bloggs wrote:
> On Saturday, February 17, 2024 at 10:46:47 PM UTC-5, Anthony William Sloman wrote:
> > On Sunday, February 18, 2024 at 3:53:24 AM UTC+11, Fred Bloggs wrote:
> > > On Friday, February 16, 2024 at 10:51:59 PM UTC-5, Anthony William Sloman wrote:
> > > > On Saturday, February 17, 2024 at 5:44:50 AM UTC+11, Fred Bloggs wrote:
> > > > > On Wednesday, February 14, 2024 at 10:32:40 PM UTC-5, Anthony William Sloman wrote:
> > > > > > On Thursday, February 15, 2024 at 4:06:53 AM UTC+11, Fred Bloggs wrote:
> > > > > > > On Wednesday, February 14, 2024 at 2:57:10 AM UTC-5, Anthony William Sloman wrote:
> > > > > > > > On Wednesday, February 14, 2024 at 4:57:38 PM UTC+11, Sylvia Else wrote:

<snip>
> > > > Perhaps, but I'm clearly better-informed than you are.
> > >
> > > Of course you're going to think that, it's an ego preservation refuge for the megalomaniac.
> >
> > Funny that you, of all people, would suggest that. Look in the mirror.
>
> Using an AI retort generator now?

It is a pretty obvious response - I didn't need artificial intelligence to come up with it. I've had to deal with the occasional self-satisfied half-wit from time to time, and their predictable indignation does get an appropriate response.

> > > > > The density function for years of life should be normal-like, a crude fit is said to be the log-normal, the logarithm of an underlying normal variate. Literature is calling it a survival distribution, which makes sense. If F(A) is the cumulative distribution (integrated ) of that density up to year A, indicating the fraction of population still alive by year A. Then the chance of an individual of age A living to age A + T, T being time interval of continued life, should be F( A + T)- F( A ). What you're after, whether you realize it or not is the distribution of T. Literature says it's an exponential distribution, and that makes no sense at all since it implies a constant death rate. If you can't compute the mean and standard deviation of that simple thing, then you have problems.
> > > >
> > > > A rather long-winded way of announcing that you don't know what you are talking about.
> > >
> > > I know exactly what I'm talking about. The fact of you saying it's long winded goes to show how weak is your so-called analytical thinking.

It was essentially meaningless word salad. It certainly didn't address Sylvia's question.

> > > > > Writeup by some blithering person:
> > > > >
> > > > > https://users.stat.ufl.edu/~rrandles/sta4930/4930lectures/chapter2/chapter2R.pdf
> > > > >
> > > > > They think they're geniuses for fitting a Weibull.
> > > >
> > > > It's a shopping list of fitting functions. There's nothing in that write-up that shows the fit of an actual function to actual acturial data.
> > >
> > > It's a parameterized distribution used for fitting exponentials, and used extensively in modeling systems for reliability engineering lifetime statistics, just something else you don't know the first thing about.
>
> > That's what I said. I don't know much about it because I've never had to do that - if my bosses wanted a more reliable system, we designed one that was more reliably by design, rather than by trrying to demonstrate that what we had was reliable enough.
>
> Care to explain how you would design reliability into the design without knowing how to analytically analyze your proposed design?

Easy. You avoid parts with known failure modes. Avoiding electrolytic capacitors is a good start.

Part reliability is established by observation, rather than analysis, and to get results quickly you over-stress them - run them hot and so forth. It's crude stuff, but it keeps the military happy.

> > > > > Survival in general wiki:
> > > > >
> > > > > https://en.wikipedia.org/wiki/Survival_function.
> > > >
> > > > That is marginally better, in that it makes passing reference to real world breast cancer data, but it doesn't make any direct connection.
> > >
> > > Statistics is a tool of scientific discovery and not the science itself. Dunno what kind of childish arrested development would think it would be.
> >
> > You do seem to think exactly that.
> >
> > > > You do go to a lot of trouble to tell us that you don't know what you are talking about.
> > >
> > > You're too ignorant with an exacerbation of stupidity to make that assessment.
> >
> > Or so you like to think.
> >
> > > > The improved survival past age 80 for males and 85 for females might be susceptible to being modelled by a Weibull function - I wouldn't know. I could ask my cousin the statistician, but even though he is retired, I'd hate to waste his time on such a pointless question.
> > >
> > > It's more than just a model. It does show that beyond a critical age range the death rate becomes constant, being directly proportional to the interval of time under consideration regardless of when that interval occurs, up to a limiting age when it rapidly breaks down.
> >
> > Except that there isn't any kind of "critical age when the death rate becomes constant". The death rate depends on the environment, and it became higher when the Covid-19 virus became epidemic and went down again when most people had been vaccinated. People become more susceptible to all sorts of fatal conditions as they get older - in the US this includes not being fast enough on your feet to get away from people who have gone postal.
> >
> Another demonstration of your inability to think analytically. When confronted with externality of that nature, techniques of 'adjustment' have been developed to eliminate the craziness so as to derive the underlying constancy of whatever phenomenon it is you're studying.

So you can ignore what's actually going on?
> The mortality in any given year, despite comprising reams of data, is conceptually a single data point. So deriving a 'baseline' as they call it will require a multitude of those data points, years. You don't just say 3 years does it like that dumb Australian government page lets on they do. The analytical way is to compute numerical variation in the data and then try to predict by how much it will corrupt the estimate of baseline. That way you end up with a range of baseline with a well-controlled confidence associated with it. Excess will be based off the mean of the resulting baseline. Maybe it takes 10 years, who knows, and nothing says the sample number stays constant.

The mortality in any given year is a lot of people dying for a whole range of different reasons. Declaring it to be a "single data point ignores that reality.
Taking the 3-year number as a baseline is dumb.If the three years included the Covid-19 pandemic, it's even dumber.

If you decide that everybody ought to die at precisely evenly spaced intervals, you can assess the Allan variance between the times at which they do die. Nobody sane would.

https://www.abs.gov.au/statistics/health/causes-death/provisional-mortality-statistics/latest-release

The death rate clearly varied through the years in question. More people die in winter - June to August in Australia - and it is worth thinking about why. You don't seem to want to.

> > > You're too much of lightweight to understand any of that, so go ahead and call bullshit- the refrain of ignoramuses.
> >
> > Or in this case, the refrain of somebody who does recognise bull-shit when he sees it.
>
> It's all relative to the use the results will be put. Your simian intellect isn't picking up on that.

Actuarial data is used by actuaries to calculate life insurance premiums. That's how it started, and the data is now applied to a bunch of other applications, mostly in public health, but that original application shapes most of the thinking about the subject - not yours, because you aren't thinking..

> > <snipped total waste of bandwidth on cut-and paste>

> > > Total waste of time to post that, you and numbers don't get along.
> >
> > We don't cooperate. I use them - which doesn't demand anything of them except their passive existence.
> >
> > > I'm finding the business pages have the best explanations for statistical principles. They do the best job of making real sense of it. The 'nurd' pages are mostly jackass-inine factoid regurgitators. The nurds are used to being confused.
> >
> > They keep it simple for the intellectually unambitious. At Melbourne University the brighter students studied economics and the dumber ones did business studies.
> > That means leaving out most of the interesting stuff.
>
> Laughable assertion. What you call interesting is probably extraneous conversation that has nothing at all to do with actually applying the methods.

How you you know? You haven't shown a hint of any consciousness of how the methods are applied or of the questions the statisticians might be trying to answer.

--
Bill Sloman, Sydney