Any good advice to someone educated as a philologist, who needs to discuss the language of statistics (and has listened to some lectures and thinks she can kinda-understand the philosophy, way to develop to understanding the formulae of course) and needs to widen her philological horizons - texts, good sites? Dirty jokes and black humour also welcome.
P.S and yeah, I do know the big lie, huge lie and statistics-thing
zilch? Actually I'm serious. The expertise around here has helped me more than once. Sometimes inspired, sometimes direct links. Thanks for all the think-tank for that, seriously.
Locate a book called "How to Lie with Statistics".
It's a gem, and not only gets across all you need to know about statistics but lays bare all the tricks used by Governments, marketeers, PR Agencies and so on to present distorted statistics without appearing to.
It's was my Bible since the 1970's, but I can't just lay my hands on it now; I'll keep looking.
It's not out of date. All the things they were doing then they are doing now, with little that's new apart from the use of computer graphics instead of paper charts and pictures. Oh, and computers now produce figures with totally spurious "accuracy", eg 15,604.28, when the reality is "somewhere betwen 10,000 and 20,000".
This trick is often used to disguise the use of a tiny sample in a huge population. "83.33% of women say that our shampoo is best!" they cry, when the sample was 18 women in the street outside the advertising agency, and the questions was "Please tell us, to get a prize, if you think that our Costicaseed Shampoo is best".
yep. I was baffled with how the scale and grids and computer graphics can make things look different for sure! I'm after good texts/sites that would help a non-native speaker interpret their graphs (seriously later), and black humour would certainly help to remember things.
The Tiger that Isn't by Michael Blastland and Andrew Dilnot, Published by Profile Books (ISBN 978-1-86197-839-4) showing how to see through the statistics quoted by 'authorities' on a subject who are using numbers to confuse or mislead.
Fascinating book - I must read it again, I've forgotten a lot of it (but don't read into that anything about the quality of the book; I've just got a very bad memory).
Quote:
75% of statistics are made up
I thought it was nearer 87.5% (or maybe that's my memory playing up again)
GG
Last edited by Groundgripper; 21st Oct 2012 at 18:58.
Location: A civilised little County..with a bit of eccentricity to boot
Posts: 678
Just mention the word..variables....you can have a lot of fun with this word when meeting the disciples of statistics and those who regard stats. as being the definitive answer to support their ( usually ) flawed presentations.
If they give a percentage, find out what the actual numbers are, and v.v.
Look at the axes on graphs; are they linear and do they begin at zero? If not, redraw them so they are and do. The result will be much less impressive.
and most importantly, what was the question asked? Are British Gas(a) Average (b) Brilliant (c) Cracking? Polls show 100% of customers think british gas is at least average
Best example is the fictional Pravda report of an Ambassadors' race, stating the Russian came second and the American was second to last. The New York times reported it as the American beating the Russian in a two-man race.
For bonus points, find out what a standard deviation is, and how sample size affects accuracy.
Lastly find how any poll was conducted. Most of the Industry-Standard techniques are scientific rubbish. Phone polls only get people who have landlines generally. Personal Surveys are usually done in a shopping mall on a weekday - not exactly a fair cross-section of society.
I would recommend any good book on how to lie with statistics, since this is how most people use them.
Edit: Capot's example is in fact spurious precision. It would be worthwhile learning the difference between accuracy and precision.
Last edited by Fox3WheresMyBanana; 21st Oct 2012 at 20:04.
Sorry probes, you'sd ta have a really good "simple" guide, but sold it on a car boot sale, and can't remember the title of it.
It sounds like you need a bit of a "handbook" that explains all the technical terms, from p-values to validity.
The psychology lot have spent former years getting beaten up for lacking robustness in their experiments, latter years trying to defend their science, so speaking to your local dept may help.
Without wishing to be to be too flippant, there are two ends of the scale..... top dollar are multicentred randomised large scale controlled trials, at the other end is someone like me giving their opinion. Everything else falls somewhere in the middle. (Which is why you need the guide)
The BMJ released a good guide a few years ago, will try and find it when I get home
A number of people here seem to be equating statistics with polling
The real use of statistics has a more serious role. For example in manufacturing: sampling the products to determine whether they are in an acceptable tolerance/quality range.
Many, many years I took a course on statistics. The lecturer showed an OHP slide showing the numbers of people admitted to lunatic asylums between 1930 and 1950. He then showed another slide showing the numbers of wireless licences issued between 1930 and 1950. He said that just because the two graphs appeared to be identical did not mean that there was correlation between the two!
).
