Damn lies and statistics….

[twentysix by twentyfive]

View attachment 607894

And there's more >>>> https://www.tylervigen.com/spurious-correlations

Yes - seems to be a measure of correlation up to and including 2007. But to my eye it's an anti-correlation in 2008 and 2009. But I'm just using the Mark I eyeball and not proper statistical analysis...............

 

[twentysix by twentyfive]

Statistics prove that figures can show anything.

The lack of understanding of the data collection methodology and the following misuse of the "statistics" can show anything. Properly controlled measurements (data collection) and analysis is very powerful in revealing truths. These truths may be limited by the scope of the measurements and analysis. That's where much misunderstanding and misuse comes in.

 

[Fab Foodie]

That example shows me that journalists are at least as bad at reporting numbers as the rest of us.

(I also think that a lot of people THINK they're good wiv data, because they're good at arithmetic, and/or can drive Excel. The science of statistics is often still alien to them ...)

I passed a module in statistics and experimental design and that was enough for me ro know statistics isn’t my thang!

 

[twentysix by twentyfive]

I passed a module in statistics and experimental design and that was enough for me ro know statistics isn’t my thang!

Have you still got the Stats text book?

 

[Archie_tect]

My son nearly failed his Marketing Masters- though his tutor coached him in the statistical analysis for his dissertation, the other lecturer who marked it fundamentally disagreed with his assertions... proving there's absolutely no objectivity in statistics

 

[matticus]

My son nearly failed his Marketing Masters- though his tutor coached him in the statistical analysis for his dissertation, the other lecturer who marked it fundamentally disagreed with his assertions... proving there's absolutely no objectivity in statistics

sounds like an outlier to me ...

 

[MichaelW2]

My alarm bells ring when some report mentions "percentage increase" without actual numbers. It is a good way to hide a small increase in a small number.

 

[winjim]

I passed a module in statistics and experimental design and that was enough for me ro know statistics isn’t my thang!

If I can graph it and get a straight line, it's probably fine. If I can't then I just log transform it until I can.

Fortunately for my dissertation the disorder I'm studying is so rare and the analyte so novel that statistical analysis is all but impossible.

So guess what 'anecdata' bores, I'm using case studies. And no, correlation does not equal causation, but association can be a powerful tool even without a demonstrated causative effect...

The power of n=2. Makes your write up a bit more manageable anyway.

 

[ColinJ]

My alarm bells ring when some report mentions "percentage increase" without actual numbers. It is a good way to hide a small increase in a small number.

Or to exaggerate the significance of a big increase in a small number...?

"The number of positive test results has increased by 50% in one day..."

Yes - from 2 to 3!

 

[gbb]

As my grandaughter is reaching the stage in life where she starts to form her own opinions, on the threshold of 6th year (I gather it's not called that anymore ?)... I tell (suggested to) her....
Form your own opinions.
If you're listening to any advice, adverts or particually anything political, ask yourself who is paying for or backing that advice or advert ?....it will almost certainly be loaded.
Dont automatically believe what anyone tells you, do your own homework.

 

[Fab Foodie]

Have you still got the Stats text book?

Somewhere! It’s green….

 

[Fab Foodie]

In my bucket chemistry world of applications (rather than proper R&D), we rarely have the resources to set-up every test that we might like (each trial is multifactorial), so we have to be smart when planning. One question I constantly stress to our newer and less experienced members is to think about exactly how you need the outcome to be (not cheating!), does the plan really tell the story, show the trend or prove/disprove your ideas? It sounds obvious, but not always!

 

[newfhouse]

This was a government minister giving evidence at the Education Committee. I will happily concede that it’s unlikely to be a left/right problem (so not a political point!), more an example of a failure of education.

 

[MartinQ]

My son nearly failed his Marketing Masters- though his tutor coached him in the statistical analysis for his dissertation, the other lecturer who marked it fundamentally disagreed with his assertions... proving there's absolutely no objectivity in statistics

Sounds he should have put another layer on his analysis and assigned priors to his assumptions, then used Bayes.
Then he could have got it wrong rigorously.

Back to the OP, though a slight detour from spurious correlations, one of my favourites is simpson's paradox
https://towardsdatascience.com/simpsons-paradox-and-interpreting-data-6a0443516765
Gets you thinking a bit ..

 

[Ming the Merciless]

Who am I to dis a Brie
I cheddar the world and a Feta cheese
Everybody's looking for Stilton...

(by Cyraneau de Beargerac)

Hold your cheese up, keep your cheese up, movin' on
Hold your cheese up, movin' on, keep your cheese up, movin' on
Hold your cheese up, movin' on, keep your cheese up, movin' on
Hold your cheese up, movin' on, keep your cheese up