Patters wrote:

I don't know what this means. Just give me one example of a well-used test statistic which is so flawed. No-one would take averages seriously in this extreme example. And more to the point - this hardly relates to the real world. Strikers don't score 40 goals in one game and 0 over the other 37. If a striker has a goal-a-game average, it's usually because they've scored regularly. And if their goals do come in patches, this will be noticeable.

It's obviously not a real test, I was just giving an extreme example of how statistics don't take the human factor into account enough.

Patters wrote:

This is often a criticism of economic analysis: 'too many unrealistic assumptions'. The reality is that - unless particular distributions and assumptions are considered completely appropriate (and they can be in large samples) - nonparametrics methods are used which set out to impose no conditions on people acting rationally and, well, minimal assumptions in general. Estimation and inference is completely based on what actually happens. Whether humans react irrationally or rationally, it doesn't matter. All that shows up is what they do. And when you have repeated observations over many individuals and/or time, the accuracy of estimation can be powerful.

(Sigh) You've forced me to quote Rumsfeld.

There are known knowns
There are known unknowns
And then there are unknown unknowns.

quote='Biggus' pid='255146' dateline='1361120135' You've forced me to quote Rumsfeld.

There are known knowns
There are known unknowns
And then there are unknown unknowns.
[/quote]

If there are unknown unknowns, they show up as systematic or nontrivial errors. Treating them as nuisance parameters there are many ways to account for them - even if you don't actually observe what they are. i.e. you turn them into known unknowns. The underlying principle of estimation/testing is this: there is some metric that looks to minimise the distance between what you think should happen (based on what your model says...and, as I've said, this is often data-driven and imposing minimal assumptions) and what actually happens. If the errors are not insignificant, your model is rubbish and it'll never be used. If it works the unknown unknowns are not worth knowing about.

You guess in other words as they are simply outside the parameters of experience, they have a bit in common with religion then.
Both inventions of humanity having no independent existence, trying to bring order into chaos.

General Mirth wrote:

I was not trying to be rude. You've genuinely not grasped the fundamentals of what statistics is, how it's used, what purpose statistical testing is and therefore any further debate on an internet forum cannot and will not address your concerns. At least, I can't. Patters and Kel are clearly more clued up on the subject.

But seriously, I didn't invent statistics and it does work in several scenarios and disciplines otherwise society wouldn't exist in the way that it does. Just understand that statistics is what we use to correct bias which you rightly pointed out is a human trait.

Neither was I, age does give you certain advantages however as you will find out one day.
I've talked to a young fit man one night, only to learn that he died as he headed the ball the next day, I know a woman who has had cancer and a heart attack yet still happily smokes and drinks whiskey well into her 70's, statistical anomalies to be sure but real people.
Nothing is certain nothing is knowable anything can happen....

Biggus wrote:

Neither was I, age does give you certain advantages however as you will find out one day.
I've talked to a young fit man one night, only to learn that he died as he headed the ball the next day, I know a woman who has had cancer and a heart attack yet still happily smokes and drinks whiskey well into her 70's, statistical anomalies to be sure but real people.
Nothing is certain nothing is knowable anything can happen....

And yet, the health insurance market works across the world without collapsing in on itself.

Three pages on and you still haven't grasped the basic concept behind probability.