There are many great writeups on E2 about statistics, and I won't reiterate all that information. I'm only here to repeat the theme, so you all don't forget it: STATISTICS ARE LIES. ALL OF THEM.

Anything posing as a fact that has a percentage of people in it is a lie. They interviewed twenty rednecks to come up with that number, which supposedly represents the entire country. Anything posing as fact that attempts to catalog the opinions of large groups of people is a lie. Neither you nor I have an opinion that can be accurately stated in multiple-choice; not that they interviewed you or I to get that statistic.

Don't believe anything they say about mass marketing, customer demographics, trends, political opinions, ANYTHING. Those are companies talking, and they're either SELLING that information for money or flat-out producing it for money. Those are the demographics, trends, and opinions (of politicians, not people) that they want you to have. It's all lies. All of it. And they'll do anything to make it sound like facts, since people are so thirsty for facts lately.

An example: Detroit city is now officially recorded with the Census Bureau as having less than 1 million citizens, for the first time in a long time. In the address list the C.B. gave to the City, they showed ONE ADDRESS for Harbor Town, an area with a known population of over one hundred thousand--more than almost all of the suburbs. Why miss so many in Harbor Town? Oh, well, THEY'RE POOR, that's why.

So just do me that favor: Never ever EVER trust a statistic, especially if the guy telling it to you is obviously a bad guy. My more general rule for myself is never trust the truth coming from an evil person, but you can just stick with avoiding statistics if you don't like dubbing people "Evil"--it doesn't bother me tho!


P.S. Regarding "scientific ideals": If I didn't know science, I wouldn't think statistics were evil, and no, I meant all of them. THE WAY STATISTICS ARE DONE TODAY involves some of the world's most screwed-up science and procedure. Most statistics that are widely disseminated have been so tampered with and seeded that they have a better chance of being dead wrong than they do of being partially right. Go read something about how they get statistics, and then come back and tell me it's at all "scientific". --PD

P.S.S. Just cause I have to: I'll put my cynicism in the way of anything I like, at least until I meet something that can even get through it.

I'd be happier if the title of this node was "96% of Statistics Are Lies, With A Margin of Error of 5%".

Some of us would hold that the set of 'all statistics' is the Truth. The mathematical representation of what happens is about as close as you can get to reality on paper.

I think you actually mean 'people use statistics to support their lies'. Statistics can be used to emphasis one piece of information; if this information is given in an misleading context, you can mislead most people. In the proper context, a statistic is true, but that context is all the other statistics, with emphasis on none.

Of course, you could also just make up something and call it a statistic, but it's not, it's just a lie. This is no different from calling a bottle of bubbly water a fusion reactor; you can call it that all you want, but it's just plain not. If a statistic doesn't represent a real account of real events, it's not a real statistic.

The objection "All statistics are lies" is not made out of a rejection of the scientific method; it is made out of a rigorous application of the laws of reason upon which th escientific method is based.

I, myself, have had a part in gathering statistics for marketing and governmental purposes. I was a telephone surveyor. The principles of telephone surveying are similar to those of door-to-door surveying, mail-in surveying, and pretty much any other kind of surveying you can imagine.

And yes, I can tell you quite assuredly, that all statistics that have to do with society and human behaviour are lies.

First, there's the "little things" -- prompting the witness, false dilemmas, complex questions, ambiguous terms, hasty generalisations, selection of unrepresentative samples, false analogies, slothful inductions, fallacies of exclusion, and other little things that the authors of surveys deliberately do to create a sample which apparently favours their visions. Need I go on here? I've seen these in action. And baby, it ain't science.

Then there's the apathy of the surveyor. Surveying is button-pushing and babbling off a script. A monkey could do it, if a monkey could talk. Do you think the average surveyor really gives a hot damn about whether or not they create a quality study? Hell no. Do you think the average surveyor gives a flaming damn about participating in the "scientific method?" No way. Do you think they're above just making up answers because it would be easier than recording all the qualified and subjective answers a respondent gives? Unh-unh.

But then there's the foolish, naïve trust that the creators of the survey place in the respondent. Respondents love to lie! They lie about their age, they lie about their shopping habits, they lie about their voting habits, they lie about everything. Lying is more fun, and socially easier, that simply saying, "None of your business."

Not to mention the analysis. Statistical analysis, when done improperly, is spectacularly improper, and rife with logical fallacies. This is especially true of social statistics, which are so much more abstract and therefore more easily mangled, distorted, and completely fabricated.

All social statistics are twisted to suit an agenda. Those with the agendas are those who most encourage the false hope that statistics are somehow science -- especially social statistics. That is why statistics should be taken with very large grains of salt, if they're not simply ignored from the start. Statistics try to fit everyone into nice little categories; but human beings have a natural tendency to defy categorisation.

Yow! First all statistics are wrong, then all statistics that involve percentages of people are wrong, then statistics that come from evil people are even more wrong, and finally most statistics are more likely to be completely wrong than partially right. Does this mean that Derek Jeter didn't bat .339 this past season?

Ok, there's a really good point behind all this that I think I ought to address before I succumb to the urge to run amuck and node All Rhetoric is Lies!, All Fiction is Lies!, and All Opinion is Lies!.

Freshman English classes often focus on rhetoric. They attempt to teach what makes a good argument, what makes good evidence, what's a persuasive way to build an argument, and so on. Statistics aren't much different from rhetoric. They're a form of quantitative argument built on top of probability theory. In English class, they teach you how to differentiate good rhetoric from bad rhetoric, and in statistics class (well, advanced ones anyway) they teach you how to differentiate good statistics from bad statistics.

In rhetoric you have evidence and in statistics you have data. If your evidence is dubious or your data are crap, so are your results. This is the point that Cletus focused on when he described surveying techniques. Most of the statistics that people are familiar with come from surveys and, as Cletus pointed out, there are gobs of ways to do it badly. Since popular publications omit methodology (Who the hell wants to know how a survey was done? Gimme the results, dammit!), it's unlikely you'll be able to catch any problems here.

In rhetoric you have logical fallacies (click on it to see a great metanode) and in statistics you have statistical fallacies. The statement "Never trust the truth coming from an evil person" is an incitement to argumentum ad hominem. When a researcher fits a line to nonlinear data, he's abusing linear regression. Again, popular publications tend to omit such details, so you won't be able to catch mistakes here.

Statistics allows an additional pitfall: misinterpretation. Seeing a correlation and assuming cause-and-effect is a big example. In The Mismeasure of Man, Gould cites another one, namely ascribing physical substance to the results of factor analysis. Luckily, you can sometimes catch this kind of thing in since it's the point where a writer states a statistic them jumps to a conclusion.

To me, the lesson is that statistics are often misused (either intentionally or accidentally) in making a point and that without knowing how those statistics were created, you can't say whether they support the argument, in exactly the same way that the statment "Most statistics...have been tampered with..." doesn't give any clue as to whether it was revealed by God or just a guess.