60 Percent of the Time, It Works Every Time

The national post? POS.

Britney Spear's Speculum: 60 Percent of the Time, It Works Every Time

It smell like Bigfoot's dick ---how did she know that?

Riddle me this, prove that God or Allah or any other sky being exists.

Show your work.

My father used to have a saying: "Figures don't lie, but liars can figure."

Unfortunately he used to say that any time anyone produced any scientific data whatsoever.
I guess I'm saying that my father was a stubborn old coot.

ArkAngel: My stats professor once showed the class that ice cream caused crime. He then pointed out how increases in ice cream consumption rose and fell at the same times as criminal activity.

Isn't that because crime rates go up in the summer time which just happens to be the period of highest ice cream consumption?

/honest question

FTA "Citing his own "myside bias," otherwise known as confirmation bias, or the tendency to favour ideas that fit with one's settled positions, Prof. Klein Phil Jones of CRU now admits that, according to the data he used, the ignorance he attributed to the left Deniers is also true of the right Alarmists, and so the headline should have been less dramatic, something closer to "Nobody Understands Economics Climate Change: Study."

The problem, as Prof. Klein Phil Jones of CRU puts it, was the hidden bias in his own use of the data, and in the decisions he made about how to analyze Adjust it."

Wow, this study is useful in so many fields!

RodneyToady: This is why there's peer review, though. No respectable journal would let something like this get published. It's like saying that all cookbooks are suspect because someone created a recipe for lemonade that contained a gallon of chocolate ice cream as an ingredient. Yeah, it was written, but nobody would take it seriously.

Oh, don't be so sure... [link]

I can get you any result you like. What's it worth to you?

wrenchboy: Riddle me this, prove that God or Allah or any other sky being exists.

Show your work.

Statistically, people are likely to believe in God. People are statistically unlikely to believe in a lie. Therefore, God is statistically probable.

doglover: The fact that they're unreliable?

Statistics are only as unreliable as your sampling method. I'm not convinced that you actually know what statistics are. Hint: when you read a newspaper article about some new study that shows X% of some population does Y, you're not reading about statistics. You're reading about a journalist with no understand of statistics trying to interpret a study and reaching a conclusion that isn't actually found by the study.

Statistics aren't needed if the number of data points is lower than the number of fingers on your research team. Otherwise, you're going to perform a statistical analysis on your data set.

Fonr anyone looking for a good read on debunking stats, experiments and pseudo-science I strongly recommend 'Bad Science' by Ben Goldacre. He tears homeopathy to shreds, amongst other things. Very good read:

http://www.amazon.co.uk/Bad-Science-Ben-Goldacre/dp/0007240198

He's also on the twitters.

Margin of error for that 68% is 67%.

ArkAngel: My stats professor once showed the class that ice cream caused crime. He then pointed out how increases in ice cream consumption rose and fell at the same times as criminal activity.

If this wasn't a cautionary example to demonstrate that correlation does not prove causation, I am angered by the quality of statistical education you received.

poot_rootbeer: ArkAngel: My stats professor once showed the class that ice cream caused crime. He then pointed out how increases in ice cream consumption rose and fell at the same times as criminal activity.

If this wasn't a cautionary example to demonstrate that correlation does not prove causation, I am angered by the quality of statistical education you received.

Politics Tab?

magic_patch: Fonr anyone looking for a good read on debunking stats, experiments and pseudo-science I strongly recommend 'Bad Science' by Ben Goldacre. He tears homeopathy to shreds, .

Stupid Ben Goldacre, he made homeopathy stronger

hubiestubert: 84% of all statisticians truly hate their f*cking job

Very close, but the first thing the headline made me think of was:

64 percent of all the world's statistics are made up right there on the spot
82.4 percent of people believe 'em whether they're accurate statistics or not


Which I think are accurate statistics ;-).


/golf clap from one stoner-folk-musician fan to another

Darrell Huff just keeps facepalming

RodneyToady: An ANCOVA analysis, controlling for their father's age, showed a statistically significant but logically impossible effect: listening to When I'm Sixty-Four made people 16 months younger than listening to Kalimba.

Listening to a song obviously has no bearing on how old you actually are. This nonsensical result, they argue, was merely an artifact of flawed analysis within a scientific culture that permits all kinds of relevant details to be excluded from the final publication.

This is why there's peer review, though. No respectable journal would let something like this get published. It's like saying that all cookbooks are suspect because someone created a recipe for lemonade that contained a gallon of chocolate ice cream as an ingredient. Yeah, it was written, but nobody would take it seriously.

The problem is, when the reviewers happen to agree with the author's thesis, even peer review fails. The seminal example of this is the work of Michael Bellesiles' work in trying to prove that guns were uncommon in early America. It went against what was considered to be common knowledge. It wasn't even a very subtle thing, yet it passed peer review because it was reviewed by what I presume to be mostly liberal historians. None of them question the data or the conclusions. They read Bellesiles' paper and apparently not one of them had any serious misgivings about the paper, which *SHOULD* have rung alarm bells. So, the paper got published, and it was expanded into a book, which also got published.

Then the people in the "gun culture" started taking a look at it, because they obviously had a motive to prove it wrong, and they noted evidence not just of what might be considered honest mistakes, but also evidence of outright academic fraud. You can read the whole sordid story of it here.

The upshot of it all was that the conclusions fit the preconceived ideas of the reviewers well enough that there wasn't any true peer review: They read it, it supported their particular world view, and so they didn't look too hard at the conclusions, even though it was pretty obvious that *SOMETHING* was amiss. The claim was extraordinary, being almost the equivalent of increased ice cream sales causing increased crime, yet it was people outside the academic world that first took a look at it and said "Hey, these numbers just don't add up", and "These records that are claimed as a source were destroyed 100 years ago".

I think the larger problem is that higher education acts as a kind of filter for differing political views. This isn't a problem in the hard sciences like astronomy, chemistry, physics, etc., but in the so-called "soft" sciences like sociology or history, it can be a significant problem. I can't think of a more effective way to suppress the possibility of dissident thought within a community than how higher education works: Faculty have a say in who gets tenure, and they're more likely to pick people who think like them. They are also more likely to give better grades to students who write papers that they agree with politically, and to judge more harshly those they may disagree with. Students with differing views than the "standard academic model" either end up not wanting to go against the flow, or they just go along with it because that's what's expected if you want to get that doctorate and a professorship with tenure. It becomes a sort of self-perpetuating intellectual echo chamber, or more provocatively, intellectual incest. And like actual incest, if unmoderated by healthy doses of outside DNA, it will inevitably lead to grotesquely misshapen offspring.

LewDux: magic_patch: Fonr anyone looking for a good read on debunking stats, experiments and pseudo-science I strongly recommend 'Bad Science' by Ben Goldacre. He tears homeopathy to shreds, .

Stupid Ben Goldacre, he made homeopathy stronger

Heh.

doglover: Sid_6.7: You know what helps the most with lying with statistics?

The fact that they're unreliable?

Seriously, by their very nature statistics are just a crap shoot. A very intricate crap shoot, but you're basically saying it's not certain. Even it's more likely your penis will sprout wings and fly away after some passing geese than something not happening I've never seen 100% certainty from a statistician.

Meanwhile, the world is based on certainty. Physical properties don't change spontaneously. Our stomachs don't start churning out HF and killing us. Gravity doesn't repel things from the center of mass spontaneously. It's just not how the universe works.

So statistics are simply elaborate lies re-jiggered to pose as truth when in reality you can ignore 'em most of the time.

Quantum mechanics and random chance inherent in any system say hi.

Statistics is a pure, mathematical methodology; there's nothing wrong about statistical analysis, the methods used, or the reached conclusions.
What we have at work here is a journalistic system with no rigor, a research structure rooted too deeply in profit, and an uneducated public. Really though, does it surprise anyone? Nobody has the time or the desire to talk about the rationality of subgroup selection or justifying the use of the log-normal distribution. They just want catchy headlines, "facts" to bring up in conversation. Even management doesn't care about details, and they're the ones who spend millions and millions on research. They'll hire a consulting firm expecting to see a 15 minute presentation followed by 15 years of profits.

Harry Ovaries: 64 percent of all the world's statistics are made up right there on the spot

I said something like this once to a friend in economics and management.
He didn't get the joke.

doglover: Gravity doesn't repel things from the center of mass spontaneously. It's just not how the universe works.

So when the news broke about the expansion of the universe acellerating a while back how did the news affect your universe view? Link (new window)

Just because things have always been that way does not mean that things will always be that way...

Jobbers: What we have at work here is a journalistic system with no rigor, a research structure rooted too deeply in profit, and an uneducated public. Really though, does it surprise anyone? Nobody has the time or the desire to talk about the rationality of subgroup selection or justifying the use of the log-normal distribution. They just want catchy headlines, "facts" to bring up in conversation. Even management doesn't care about details, and they're the ones who spend millions and millions on research. They'll hire a consulting firm expecting to see a 15 minute presentation followed by 15 years of profits.

Also THIS

dready zim: So when the news broke about the expansion of the universe acellerating a while back how did the news affect your universe view?

There's no reason to think that gravity is repelling things from the center of mass. No one knows what Dark Energy is, which is why it gets that ominous sounding name.

bow: I highly doubt 84% of statisticians hate their job.

We're going to need a p-value on that hypothesis...

/68 point 3

There are three kinds of lies: Lies, damned lies, and statistics.

Attribution is contested.

I think Penn & Teller did and episode of BS on statistics and how people giving polls can manipulate them

bow: Many people don't understand statistics, so they think it's a bunch of shiat. It's not. I have one semester of graduate school left, and my major is biostatistics. There is a bunch of garbage out there, as far as statistics, but like someone already said this is why we have peer-reviewed journals.

This!

Make probability and such shiat high school requirement and teach actual statistics to everyone. Yes, university already have requirements of statistics for MBAs, engineering, finance, biology, psychology, nursing and so on, but most people who have taken a statistics class don't remember a single thing.

Even PhD in life sciences that write papers giving statistical proofs don't really understand statistics, they just use a sequence of steps to get a set of numbers and if the numbers are between so and so, it's proof! They don't even really know what those number really signify.

This is made worse by journalists who use 74% of this said this and that which is so meaningless.

bow: Jobbers: Statistics is a pure, mathematical methodology; there's nothing wrong about statistical analysis, the methods used, or the reached conclusions.
What we have at work here is a journalistic system with no rigor, a research structure rooted too deeply in profit, and an uneducated public. Really though, does it surprise anyone? Nobody has the time or the desire to talk about the rationality of subgroup selection or justifying the use of the log-normal distribution. They just want catchy headlines, "facts" to bring up in conversation. Even management doesn't care about details, and they're the ones who spend millions and millions on research. They'll hire a consulting firm expecting to see a 15 minute presentation followed by 15 years of profits.

This.

Many people don't understand statistics, so they think it's a bunch of shiat. It's not. I have one semester of graduate school left, and my major is biostatistics. There is a bunch of garbage out there, as far as statistics, but like someone already said this is why we have peer-reviewed journals.

I highly doubt 84% of statisticians hate their job. I love what I do. If you don't like/believe in statistics, that's fine. Assume what you want, but I'll actually be proving or disproving data with mathematics and science. I'll be enjoying what I do and making good money at the same time.

This.

Plus: The only way we know something in science (hard, soft, whatever - anything where numbers are involved) is through statistics. You never get the exact same numbers twice, even with the same setup. So you either choose the numbers you like best (i.e. fudge the data) or use statistics (e.g. take the mean), and then try to deduce the law or other relationship.

ArkAngel: My stats professor once showed the class that ice cream caused crime. He then pointed out how increases in ice cream consumption rose and fell at the same times as criminal activity.

So what you are saying is we should never listen to your opinions ever again, since you believe that your stats professor could show such a thing. Good to know :P

This article is a pile of garbage that I could only sift halfway through. The bias in many of these studies is well known by anyone in the field or with basic statistical knowledge merely by glancing at them.

Any psychologist will tell you that a study has to be done, redone, done by others, done some more, done a little differently, then done to death before you can walk away drawing real conclusions. Any psychologist whose primary interest is "proving" something about a large group of people with a target audience of idiots browsing the internet or reading a popular magazine is generally going to be presenting a poorly done, never reproduced study that is then doubly slanted by the journalist.