The Naked Scientists

The Naked Scientists Forum

Author Topic: How are statistics abused?  (Read 5932 times)

Paul Anderson

  • Guest
How are statistics abused?
« on: 06/09/2008 15:35:54 »
Paul Anderson  asked the Naked Scientists:

Hi Chris and team,
I presume you accept statistics as part of science?
 
A chap rings me up and wants me to take part in a phone survey. I say "No thanks" and hang up. He rings a second person and gets a lonely old woman or man who never gets a chance to talk to anyone and they take part in the survey. He rings a 3rd person who asks what the survey is about, decides he has a strong view on it and takes part.
He rings a 4th person and as it is teatime and the housewife if busy making tea, she declines. He rings a 5th and the person says they cannot understand his Indian or Filipino accent and they hang up.
He rings a 6th and a small child says that "Daddy has not come home from work yet" and hangs up. He tries during the day and rings a 7th person who is unemployed and doesn't want to be employed and is living off the dole, and takes part in the survey.

The results are given in the radio news and they say it has an error margin of 3 to 4 per cent! How can they say that?
 
If the poll taker rows to an island originally populated by 10 people, 5 of whom are now dead or have left the island, so he interviews the remaining 5, the poll results might be considered complete in that every person on the island has been interviewed. It now a matter of breaking it up into male and female, ages, occupations, etc.
 
Unfortunately I believe statistics can be doctored to show anything.
Is the cup half full or half empty? You can put whichever spin you want to that. The Democrats have managed to keep the cup at least half full. The Republicans have half emptied the cup!  Squandered half?
 
One of the annoying things I see in the media is when they say they (scientists employed by the government) have not been able to prove that x is detrimental to your health.

That does not mean it is okay. It may just be that the scientists have not been researching it properly.  Smoking, DDT, Agent Orange, exposure to radiation, asbestos...

Regards
 
Paul
What do you think?


 

Offline petrovitch

  • Jr. Member
  • **
  • Posts: 19
    • View Profile
How are statistics abused?
« Reply #1 on: 07/09/2008 05:56:30 »
This subject was covered extensively in a book published in 1954 by Darrell Huff entitled How to Lie with Statistics.  After reading this book you will find examples of how someone has lied (or unintentionally misrepresented the facts) in almost every newspaper article you read.  To conserve space newspapers often print graphs that do not begin with zero thus illustrating trends, cycles, and seasonal movements in data that do not exist.

Another great title is Innumeracy by Douglas R Hofstadter.  The term was coined to represent how little the general public knows about numbers.  It is a play on words referring to an illiteracy with numbers.  It cites several examples of lawyers using statistics to prove their case when, in fact, the statistics are meaningless or even disprove what they intended to prove.

I can give lots of examples if you wish.
« Last Edit: 07/09/2008 06:00:44 by petrovitch »
 

Offline Bored chemist

  • Neilep Level Member
  • ******
  • Posts: 8655
  • Thanked: 42 times
    • View Profile
How are statistics abused?
« Reply #2 on: 07/09/2008 09:41:51 »
Respectable polling organisations are aware of the broblem and they take steps to get round it. For example, they know roughly what fraction of the population are unemployed- say it's 10%. They also realise that more unemployed people are likely to answer the phone so when they have collect their data they ask if theperson is emplyed or not.
When they get the data it might show that 20% of the people poled are unemployed. They throw away half that data so what they are left with is a better representation of the population as a whole. They do similar things for age, sex, and other groups.


Throwing data away like this is expensive so not all polling organisations do it. It's just another example of the fact that you get what you pay for.

Incidentally, Darren Huff's book may be a bit old, but it's a very good introduction to statistics and very readable. I think it should be compulsory reading at school. The same is probably true of innumeracy- if everyone were fammiliar with these books there would be a lot fewer fraudsters out there.
 

Offline petrovitch

  • Jr. Member
  • **
  • Posts: 19
    • View Profile
How are statistics abused?
« Reply #3 on: 07/09/2008 20:29:57 »
Correction.  The is the author of Innumeracy is John Allen Paulos -- not Douglas R Hofstadter. He made comments about the book on the back cover.
 

Offline JimBob

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 6564
  • Thanked: 7 times
  • Moderator
    • View Profile
How are statistics abused?
« Reply #4 on: 08/09/2008 03:37:28 »
This subject was covered extensively in a book published in 1954 by Darrell Huff entitled How to Lie with Statistics.  After reading this book you will find examples of how someone has lied (or unintentionally misrepresented the facts) in almost every newspaper article you read.  To conserve space newspapers often print graphs that do not begin with zero thus illustrating trends, cycles, and seasonal movements in data that do not exist.

Another great title is Innumeracy by Douglas R Hofstadter.  The term was coined to represent how little the general public knows about numbers.  It is a play on words referring to an illiteracy with numbers.  It cites several examples of lawyers using statistics to prove their case when, in fact, the statistics are meaningless or even disprove what they intended to prove.

I can give lots of examples if you wish.

I found this book in 1974 - consider it one of the keys to my success!

Small Book, big change in the way you look at things and do things yourself - And best read if you have good, high moral standards for yourself.

 

Offline Soul Surfer

  • Neilep Level Member
  • ******
  • Posts: 3345
  • keep banging the rocks together
    • View Profile
    • ian kimber's web workspace
How are statistics abused?
« Reply #5 on: 08/09/2008 12:45:01 »
Statistics is a very valuable scientific tool and without it we would not be able to discover many important relationships. 
However the fact that numbers appear to be linked does not guarantee a causal relationship and that needs very careful investigation and the analysis of alternative processes.
Unfortunately vested interests often use statistics to "prove" their points and the general public can be very gullible
 

Offline petrovitch

  • Jr. Member
  • **
  • Posts: 19
    • View Profile
How are statistics abused?
« Reply #6 on: 09/09/2008 02:36:09 »
Multicollinearity is a common error in articles trying to prove cause/effect.  Many journalists confuse correlation   with cause/effect.  The textbook examples is the misinterpretation of cause/effect given a positive correlation between the number of sun spots and gross national product.  Nikolai Kondratieff discovered 50 year business cycles, but there is no cause/effect.

Another common error that I see made daily by economists and journalists alike is the abuse of building a model under the assumption of ceteris paribus. With all things held equal a single variable is modified and examined for positive results in a business model. 

This can lead to false security in production and financial models; in the real world all other things are not constant. 

For example, recently in the grain market pessimistic and optimistic models forecasting a change in price did not fairly estimate the effects of the rate of exchange (currency) and the effects that fuel would have on basis.  A non-budget item, like basis, had overwhelming effects on profit margins.
 

The Naked Scientists Forum

How are statistics abused?
« Reply #6 on: 09/09/2008 02:36:09 »

 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums
 
Login
Login with username, password and session length