PDA

View Full Version : "How to lie with statistics"



stalkervision
11-21-2008, 12:25 AM
Seems like a perfect book for these forums.. http://forums.ubi.com/groupee_common/emoticons/icon_biggrin.gif

http://ecx.images-amazon.com/images/I/51HRGNPNEYL._BO2,204,203,200_PIsitb-sticker-arrow-click,TopRight,35,-76_AA240_SH20_OU01_.jpg

http://www.amazon.com/exec/obidos/ASIN/0393310728/dansdata

stalkervision
11-21-2008, 12:25 AM
Seems like a perfect book for these forums.. http://forums.ubi.com/groupee_common/emoticons/icon_biggrin.gif

http://ecx.images-amazon.com/images/I/51HRGNPNEYL._BO2,204,203,200_PIsitb-sticker-arrow-click,TopRight,35,-76_AA240_SH20_OU01_.jpg

http://www.amazon.com/exec/obidos/ASIN/0393310728/dansdata

Kettenhunde
11-21-2008, 05:00 AM
If you have never had a statistics class, that is a great book.

It goes over most of the items you will find covered in the begining of a formal class. How graphical representations or data sampling can be be misleading for example.

The book really helps those ignorant to the value of statistics how to properly interpret the information. All to often is not the math that "lies" it is the person who improperly interprets the results in ignorance.

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content"> More to the point, with the help of this book, you need not have blind faith in the numbers or disgustedly throw all stats away. The mathematics of statistics guarantees them to have great power, as long as you know how to interpret them correctly. You might be pleasantly surprised to find that more common sense than math is involved in this book, but the truth is most modern abuse of numbers happens well after the numbers have been calculated. </div></BLOCKQUOTE>

http://www.amazon.com/exec/obidos/ASIN/0393310728/dansdata

All the best,

Crumpp

josephs1959
11-21-2008, 07:42 PM
Kudos stakervision! LOL! Everyone (with the exception of some members here)knows that the FW190A is considered the "perfect fighter" of WW II.
Meaning it does ALL things very well if not perfect.

josephs1959
11-21-2008, 07:44 PM
The fuse is lit, I now leave you to the following explosion. LOL!

Divine-Wind
11-21-2008, 08:09 PM
http://forums.ubi.com/images/smilies/88.gif

TgD Thunderbolt56
11-21-2008, 10:42 PM
Is that something like: Liars can figure, but figures don't lie?

R_Target
11-21-2008, 10:49 PM
Some don't even need statistics to lie.

M_Gunz
11-21-2008, 11:09 PM
You can fool some of the people all of the time...
A good convincing lie is based on truth...
-and-
There's a sucker born every minute.

ShaK.
11-22-2008, 09:35 AM
Can you give an example of what is in this book?
the only reason I ask is, out of 68% of people who read this thread, 42% will not understand what it is about.

M_Gunz
11-22-2008, 10:07 AM
So it isn't a how-to for government work?

JSG72
11-22-2008, 11:27 AM
Statistics IMHO can only be of use.

Where the intervention of humans has never taken place. In other words...

If you beleive in them. You are in the hands of the Gods. And we wouldn't have to build Hadron Colliders.

Lies! Damn Lies and Statistics....

So what was the probabilty of someone coming ouy with that statement. And does it matter that this thread goes O.T.

M_Gunz
11-22-2008, 01:12 PM
Look up the first use of statistics. It stemmed a huge source of disease and suffering in London.

Statistics are also used well in Quality Control that ensures better products than otherwise at lower cost than otherwise with less
waste than otherwise.

It's a tool is all. In the right hands you get good results and in the wrong hands you get marketing.

Divine-Wind
11-22-2008, 01:20 PM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by M_Gunz:
So it isn't a how-to for government work? </div></BLOCKQUOTE>
I thought this was required reading for aspiring politicians? http://forums.ubi.com/images/smilies/16x16_smiley-wink.gif

JSG72
11-22-2008, 01:47 PM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by M_Gunz:
Look up the first use of statistics. It stemmed a huge source of disease and suffering in London.

Statistics are also used well in Quality Control that ensures better products than otherwise at lower cost than otherwise with less
waste than otherwise.

It's a tool is all. In the right hands you get good results and in the wrong hands you get marketing. </div></BLOCKQUOTE>

And for most published times states the "Bloomin' Obvious." Thems 'll be the "Good ones" then?

beNdeR__
11-22-2008, 04:08 PM
Statistics can be used to prove anything. 86% of people know that

M_Gunz
11-22-2008, 04:30 PM
There's a guy giving a talk on TED.COM who is using statistics in a good way to update old views on 3rd world countries.
It may lead to better responses in world change just for starts.

Aaron_GT
11-23-2008, 10:32 AM
From another thread:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">1. P-47: Broke. Don't work. explain to me how the F(*k a damn p-47 looses the entire tail section from 20 mm, when Bob Johnson had an entire ammo load emptyied into his his plane.
Yet in this sim a few well placed 20mm will take the entire tail section off. </div></BLOCKQUOTE>

A few well placed 20mm should take the tail off any of the planes we have. The P-47 was tough but Johnson was still damn lucky that nothing vital was shot off. Those that weren't so lucky generally aren't recorded.

Basically those that returned are a self selecting set and there is not much you can say about them in statistical terms. It's a bit like having 250 people wearing hats of 10 different designs and getting them to throw dice and if someone throws three sixes in a row you (you are likely to get only one or two managing it) declare the hat design they are wearing lucky. It may or may not be lucky, it is just that the test won't actually reveal whether it is lucky or not because it is not statistically significant. You might get some indication of hat effectiveness looking at those who threw two sixes in a row, but it might still just be random. Plus you might be failing to pick up some other association - perhaps all those wearing 'lucky' trilbies are also wearing red socks and red socks have some previously unnoticed effect on the physics of rolling dice. Or perhaps the trilby wearers have unusual dice rolling skills such that they are better at rolling sixes...

stalkervision
11-23-2008, 10:55 AM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Aaron_GT:
From another thread:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">1. P-47: Broke. Don't work. explain to me how the F(*k a damn p-47 looses the entire tail section from 20 mm, when Bob Johnson had an entire ammo load emptyied into his his plane.
Yet in this sim a few well placed 20mm will take the entire tail section off. </div></BLOCKQUOTE>


A few well placed 20mm should take the tail off any of the planes we have. The P-47 was tough but Johnson was still damn lucky that nothing vital was shot off. Those that weren't so lucky generally aren't recorded.

Basically those that returned are a self selecting set and there is not much you can say about them in statistical terms. It's a bit like having 250 people wearing hats of 10 different designs and getting them to throw dice and if someone throws three sixes in a row you (you are likely to get only one or two managing it) declare the hat design they are wearing lucky. It may or may not be lucky, it is just that the test won't actually reveal whether it is lucky or not because it is not statistically significant. You might get some indication of hat effectiveness looking at those who threw two sixes in a row, but it might still just be random. Plus you might be failing to pick up some other association - perhaps all those wearing 'lucky' trilbies are also wearing red socks and red socks have some previously unnoticed effect on the physics of rolling dice. Or perhaps the trilby wearers have unusual dice rolling skills such that they are better at rolling sixes... </div></BLOCKQUOTE>

http://forums.ubi.com/images/smilies/10.gif The german pilot was out of 20mm ammo as I recall and only had machine gun bullets left. That's how Johnson did it. http://forums.ubi.com/images/smilies/16x16_smiley-wink.gif

Aaron_GT
11-23-2008, 11:00 AM
Well my point was as much the issue of picking data and that people often argue based on a self-selecting set of outliers. But your point could be a good one too! http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif

stalkervision
11-23-2008, 11:28 AM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Aaron_GT:
Well my point was as much the issue of picking data and that people often argue based on a self-selecting set of outliers. But your point could be a good one too! http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif </div></BLOCKQUOTE>

http://forums.ubi.com/groupee_common/emoticons/icon_biggrin.gif

M_Gunz
11-23-2008, 11:43 AM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by stalkervision:
http://forums.ubi.com/images/smilies/10.gif The german pilot was out of 20mm ammo as I recall and only had machine gun bullets left. That's how Johnson did it. http://forums.ubi.com/images/smilies/16x16_smiley-wink.gif </div></BLOCKQUOTE>

There were some 20mm hits on his plane but IIRC they came from AA and nothing like a full load of those!

But that one example is used to play the rules from exceptions game again and again.
Someone might as well charge me for crossing a local bridge on grounds it was sold to them at a bar last night.

stalkervision
11-23-2008, 01:14 PM
If he had full 20mm ammo belts, Johnson and his P-47 would have been toast me thinks.. http://forums.ubi.com/images/smilies/winky.gif

JSG72
11-23-2008, 02:35 PM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by stalkervision:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Aaron_GT:
Well my point was as much the issue of picking data and that people often argue based on a self-selecting set of outliers. But your point could be a good one too! http://forums.ubi.com/groupee_common/emoticons/icon_smile.gif </div></BLOCKQUOTE>

http://forums.ubi.com/groupee_common/emoticons/icon_biggrin.gif </div></BLOCKQUOTE>

+1

There is nothing predictable about "Human Intervention".
Not from samples taken to the examples questioned to analysed to concluded. As soon as., us humans, are involved it goes T*ts Up. Just ask the Worlds Banking Community? http://forums.ubi.com/images/smilies/16x16_smiley-wink.gif

Because All statistics are based on the collators/contributers knowledge on how to Ask/answer and Evaluate. They therefore become a Tool for someone. Who has neither the common sense/education. To be able to work out things for themselves. That is to say that they have no experience or knowledge in the subject being analysed due to their preferred vocation. (They had to do something to make themselves feel important.) http://forums.ubi.com/images/smilies/25.gif

Aaron_GT
11-23-2008, 02:49 PM
JSG, just because some analyses are flawed does not necessarily mean that all analyses are flawed to the extent that conclusions based on them are invalid, which is what you seem to be implying. Heck, if you start looking at statistical mechanics then the macro world is just a statistical average. I am really not sure what axe you have to grind here.

My point was really a narrow one - you can't use a statistically insignificant self-selecting sample to argue a generality. It is often done, but a good statistical analysis will avoid this pitfall. Ideally the data collection will include as much data as possible so a PCA or similar analysis can be done.

Aaron_GT
11-23-2008, 02:50 PM
P.S. This topic might be better in the OT forum.

AnaK774
11-23-2008, 03:03 PM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by stalkervision:
If he had full 20mm ammo belts, Johnson and his P-47 would have been toast me thinks.. http://forums.ubi.com/images/smilies/winky.gif </div></BLOCKQUOTE>
Unless original whiner about that and Johnsons shooter share same gunnery skills http://forums.ubi.com/groupee_common/emoticons/icon_biggrin.gif

M_Gunz
11-23-2008, 03:18 PM
Statistics is used in the artillery quite well.

JSG72
11-23-2008, 03:31 PM
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Aaron_GT:
JSG, just because some analyses are flawed does not necessarily mean that all analyses are flawed to the extent that conclusions based on them are invalid, which is what you seem to be implying. Heck, if you start looking at statistical mechanics then the macro world is just a statistical average. I am really not sure what axe you have to grind here.

My point was really a narrow one - you can't use a statistically insignificant self-selecting sample to argue a generality. It is often done, but a good statistical analysis will avoid this pitfall. Ideally the data collection will include as much data as possible so a PCA or similar analysis can be done. </div></BLOCKQUOTE>
http://forums.ubi.com/images/smilies/16x16_smiley-very-happy.gif That without doubt is the case. But if any analysis fits in with the perpetrators agenda. Then sure as "eggs is eggs". That is the outcome that will be adopted as "Fact"

All analysis isn't flawed. I.E. The ones that are blatently obvious. To an educated mind

In other words. It is alright me spouting these arguments. But If I don't have the scientifically proven? results (Through Statistical analysis) to back me up. Well without them? I am only surmising. http://forums.ubi.com/images/smilies/25.gif

Oh! and as things are going pretty slow on this Forum. Maybes we could keep it in?

WTE_Galway
11-23-2008, 05:29 PM
The whole point of statistics (used properly) is to determine what trends actually exist in data versus appear to exist but are mere coincidence.

However for it to achieve that you need a good understanding of the methodology and how it applies to the situation at hand.

The problem isn't the statistics themselves it is that people either misunderstand and hence accidentally misuse the stats (Psychologists are notorious for applying batteries of tests inappropriately) or in some cases deliberately misuse them (journalists, marketing people and political parties are the main culprit here).



What does that mean ?? Simply this, if you are not familiar with the field and the stats used it comes down to whether or not you trust the competence and integrity of the person using them.

stalkervision
11-23-2008, 05:58 PM
I want someone here to develop a statistical program having to do with certain key words and phrases and their statistical post count pages. Then all one would have to do when someone makes a new post is type in these key words or phrases such as Fw-190/uber plane or P-51/ uber duber plane ect..and get a statistical answer to the number of pages it should generate.. http://forums.ubi.com/groupee_common/emoticons/icon_biggrin.gif

anyone got any "key phrases" we can start with? http://forums.ubi.com/images/smilies/winky.gif