But Statistics are Lies!!!

[Dangerosa]

I work for a Six Sigma company. Six Sigma (a statistical quality methodology) is supposed to be enculturated (I know, look at the bad business buzz words). This means that nearly everyone at my level and above has been through at least 48 hours of statistical training - and it isn't bad statistical training....and yet you still hear..."Statistics are lies...you know, there are three kinds of lies...lies, damn lies, and statistics." and "But I don't trust numbers."

Now, its pretty easy to lie with statistics to someone who doesn't understand statistics. But the power in knowing even a little about stats is that it keeps people from being able to lie to you with stats. What is your sample size? How did you choose your sample? Exactly what was that survey question? How accurate was your measurement system? What is your confidence interval? Is that a normal distribution? Know when to ask these questions (and understand what the answers mean) and it becomes nearly impossible for someone to lie to you. And, of course, it allows you to back up what you are saying.

I don't trust numbers either. Until I understand how we got the numbers. There is a big difference between objective and subjective information. A big difference between a machine that measures to a millimeter of accuracy (or better) and a technician with a school ruler. A huge difference between timing a task with a stopwatch 300 times and asking someone "how long does it take for you to usually do that." You can get numbers for all of this, but that doesn't mean they are good.

This is what we were all supposed to learn in class (and what was taught to us) - and yet we have two types of people driving me batty - one that won't trust a number based on a randomly selected sample of 500 with a good measurement system behind it and normal data because "all statistics are lies" and "I don't like numbers." And the other that thinks if you put a number on something ("On a scale of one to ten, how do you feel about that") it becomes good data.

And I'm not sure what drives me crazier....that these people don't understand basic stats....or they don't understand basic politics enough to pretend like they embrace it and trust it. (OK, easy, it drives me nuts that these people don't understand basic stats. It only makes me shake my head and wonder that they'd spend forty eight hours in a room, hear this talked about at the highest levels constantly, and then admit that they don't buy it).

[Bren_Cameron]

Well, you know, there are three kinds of people in the world: People who are good with numbers and people who aren't.

No, actually, I'm with you all the way. I took a single statistics course in college and have been amazed at how enlightening it was, and how useful it's been in my life. And that was just one tiny, very basic class.

[ultrafilter]

There's nothing like basic statistical knowledge to piss you off.

[whiterabbit]

I took a very basic statistics-for-psych-majors class a few years ago. I'd have to review the math to remember that part of it, but as for the rest, yes, it's VERY useful. Especially when confronted with blatant stupidity like "48% of people are for X, while 52% are against it. And there's a 4% margin for error." THAT DOES NOT MEAN THAT THE PEOPLE WHO ARE AGAINST IT ARE THE MAJORITY. DUH.

[ccwaterback]

The average temperature in Chicago is 48 degrees, bring a light jacket.

[Andros_X]

And remember, folks, 93% of statistics are made up.

[Shade]

Originally Posted by Andros_X

And remember, folks, 93% of statistics are made up.

And 134% of statistics are complete nonsense.

[kferr]

Just showing off my sig

[yojimbo]

Originally Posted by kferr

Just showing off my sig

[panto]Oh no you're not[/panto]

[CrankyAsAnOldMan]

Quantitative literacy ought to be a bigger priority. It would be harder for anyone to "lie with statistics" if people were better able to interpret and question their meaning.

FWIW, I think the press needs to be first in line signing up for any effort to beef up statistical understanding.

[Dangerosa]

I think one of the things that bugs me about this is that we really have made an effort to beef up our quantitative literacy. We are talking about people who made it through college, who we put through a week plus of stats training, who still don't understand the importance of sample size. Who can't even just trust people brighter than that to say "you don't need to understand the formulas, just understand that if you flip a coin twice and it is heads both times, you don't draw the conclusion you have a 2 headed coin. Flip it 100 times and it comes up heads, and maybe you have a 2 headed coin."

[chique]

My favorite bit of statistical horseshit:

Five or six years ago scooters (Remember those? Skateboards with steering?) made a big comeback. No one knew what they were and all of a sudden every kneebiter on the block had one. There was a little news headline in the Strib along the lines of "Injuries from scooters up 600% in the past year".

Well no SHIT. If sales of scooters goes up 1000% in a year it's only logical that injuries from them go up a little, donchathink?

Grrrr.

[ElvisL1ves]

Originally Posted by ccwaterback

The average temperature in Chicago is 48 degrees, bring a light jacket.

My favorite is "The average person has one tit and one ball".

Right, single numbers can be misleading. There is no way to avoid looking at the distribution if you want to know what's happening, and that too is rarely enough.