Awesome link, statmitter!

That was awesome.

Nice! The confidence interval that this link might prove useful in explaining p-values to those who don't understand them is 0.57 to 0.99 with a p-value of 0.0001......

Statisticians aren't normal.

Mentat: Statisticians aren't normal.

No, they aren't. They're average.

Pretty good, but I'm not sure why they don't use the term "false positive." p ≤ .05 means "less than 5% chance of a false positive," which is a fairly concise way of putting it. Laypeople can relate to this concept; if nothing else, you can explain it in terms of pregnancy (or any arbitrary medical screening) tests.

squidzilla: Mentat: Statisticians aren't normal.

No, they aren't. They're average.

LOL

squidzilla: Mentat: Statisticians aren't normal.

No, they aren't. They're average.

That's just mean.

Mentat: squidzilla: Mentat: Statisticians aren't normal.

No, they aren't. They're average.

That's just mean.

It's a standard deviation...

This guy doesn't understand that even the readers of Science have trouble understanding what's meant by 40% POP in the weather report. I really doubt that some wording change is going to affect their confidence in a study.

The liberal arts majors probably abandoned the link after the first paragraph.

OgreMagi: The liberal arts majors probably abandoned the link after the first paragraph.

not true. I read to the 2nd question or so.

/ok, so I'm getting my masters, so I'm probably above average.

squidzilla: No, they aren't. They're average.

More like bimodal. ;)

wildsnowllama: Mentat: squidzilla: Mentat: Statisticians aren't normal.

No, they aren't. They're average.

That's just mean.

It's a standard deviation...

r you sure?

Statisticians do it but end up in trials.

NeauxFear: squidzilla: No, they aren't. They're average.

More like bimodal. ;)

Farking mean.

An engineer is working at his desk in his office. His cigarette falls off the desk into the wastebasket, causing the papers within to burst into flames. The engineer pulls out his engineering pad, does a quick calculation, grabs the fire extinguisher and sprays the fire for exactly 5.2 seconds to put it out, and goes back to work.

A physicist is working at his desk in another office and the same thing happens. He looks at the fire, looks at the fire extinguisher, and thinks "Fire requires fuel plus oxygen plus heat. The fire extinguisher will remove both the oxygen and the heat in the wastebasket. Ergo, no fire." He grabs the extinguisher, puts out the flames, and goes back to work.

A statistician is working at his desk in another office and the same thing happens. He immediately grabs a piece of paper, lights it on fire and begins setting fires throughout the room. "What are you doing???" shout his co-workers, to which he replies, "I need more data points!"

musashi1600: wildsnowllama: Mentat: squidzilla: Mentat: Statisticians aren't normal.

No, they aren't. They're average.

That's just mean.

It's a standard deviation...

r you sure?

I see we're in pun mode

Statisticians, the only ones who care what 35% of people whom watch daytime television think.

/get your confounding variables out of my null hypothesis

Barakku: I see we're in pun mode

it's an Alpha thing

actually, I think it would be wise to step out of this thread before I regress into a dummy varible.

mr lawson: Barakku: I see we're in pun mode

it's an Alpha thing

ruh rho says Scooby

So p-value is a stuck up, holier-than-thou way of saying probability?

bhcompy: So p-value is a stuck up, holier-than-thou way of saying probability?

It's a specific type of probability. And it's used by everyone in all the sciences, as far as I know, so it's only as stuck up as academia as a whole is.

bhcompy: So hurr?

Yes, us intelligentsia invent fancy terms just to make you proles feel inferior.

yogaFLAME: bhcompy: So hurr?

Yes, us intelligentsia invent fancy terms just to make you proles feel inferior.

If you want a picture of human history, just picture bhcompy being slammed in the face with a math book... forever

all right, I admit, I don't get it.

a double-blind placebo controlled study shows that in disease A, treatment X reduces 1 year mortality by 50% with p
isn't it the same to say, "there is a more than 95% chance that treatment X is effective," and "if treatment X really did not work, then there is a less than 5% chance that we would get same results on repeated testing?"

according to VDG, the former is incorrect, the latter is correct, but it seems to me they're the same. please enlighten me.

sorry, let me try again...

all right, I admit, I don't get it.

a double-blind placebo controlled study shows that in disease A, treatment X reduces 1 year mortality by 50% with p less than 0.05.

isn't it the same to say, "there is a more than 95% chance that treatment X is effective," and "if treatment X really did not work, then there is a less than 5% chance that we would get same results on repeated testing?"

according to VDG, the former is incorrect, the latter is correct, but it seems to me they're the same. please enlighten me.

shanghaid: sorry, let me try again...

all right, I admit, I don't get it.

a double-blind placebo controlled study shows that in disease A, treatment X reduces 1 year mortality by 50% with p less than 0.05.

isn't it the same to say, "there is a more than 95% chance that treatment X is effective," and "if treatment X really did not work, then there is a less than 5% chance that we would get same results on repeated testing?"

according to VDG, the former is incorrect, the latter is correct, but it seems to me they're the same. please enlighten me.

The key point is that the hypothesis was that the treatment didn't work, so the p-value is only related to that. You can't use it to talk about the odds the treatment is effective.

As the interviewee said, it's kind of like the difference between "I own the house" and "the house owns me".

shanghaid: sorry, let me try again...

all right, I admit, I don't get it.

a double-blind placebo controlled study shows that in disease A, treatment X reduces 1 year mortality by 50% with p less than 0.05.

isn't it the same to say, "there is a more than 95% chance that treatment X is effective," and "if treatment X really did not work, then there is a less than 5% chance that we would get same results on repeated testing?"

according to VDG, the former is incorrect, the latter is correct, but it seems to me they're the same. please enlighten me.

This has to do with how you come about with the value of p. The thresholds are all calculated based on random chance, and can only be calculated in that way. So, when you get a result that is less than your critical p (aka alpha) threshold, you reject the null (the null being the hypothesis that the independent variable [in this case treatment X] has no effect on mortality). A rephrasing of the null hypothesis would be to say that, 'any effect seen is a result of chance alone.'

When you're calculating p, you're only trying to determine whether you should reject the null, or if you will fail to reject the null. That's it. If obtained p is less than or equal to a critical p of .05, you're saying that if you repeated the test 100 times, you believe only 5 of those times you would see results of the same magnitude because of chance.

Again, you're only able to deduce how much of a role chance is playing in the data you've collected. It may seem convincing to look at it as being wrong only 5% of the time, but that's not quite enough. Like they said in the article, if you really want to see the effects of something, you need to look at the confidence interval, and that's a whole other beast.

To restate it more simply:

Even though p (alpha) says that you'd be making a mistake 5 times out of 100, it's not saying you'd be right the other 95 times. Being right means that there is another factor at play other than chance, and so the value of that (95 in this case) is not comparable to the thresholds that have been calculated based on chance alone.

Does that make more sense? That's how I understand it, but feel free to correct me if I'm wrong. I'm also assuming they are talking about alpha when they say p, and not power. Power is totally different and you don't measure it against the same thresholds as alpha, it just is what it is. I'm thinking they confounded the variables there, elsewise p and alpha are synonymous.

For those who don't get it, an explanation as simple as I can possibly make: P-value represents the probability that you would get the same result if the vaccine were totally useless. That, however, is an independent value from the probability the vaccine is useless. If the result in the article was obtained with two placebos (say, salt water vs plain water) that wouldn't mean that salt water is probably an AIDS vaccine, because the prior probability of salt water being an AIDS vaccine is even lower than the probability of it happening by random chance.

The reason that a low P-value is valuable in this case is that we already have good reason to presume that the formulation in question could vaccinate against AIDS -- but because we don't know the exact probability, we have to rely on the improbability of other results. A p-value of 0.05 is somewhat arbitrarily chosen as the cut-off point for statistical significance.

I'm a social statistician, and an alpha of .05 is more than enough for me. If I make a Type II error, in all probability lives will not be on the line.

But for medicine, I just can't see how anything bigger than .01 is tolerable. As 006007 says, assuming there is no relationship between the drug and infection rates, if you ran 100 trials you would find a correlation at least as big as yours 5 times. That seems high to me. Now, run three trials and get a p

Ugh...i meant a type I error.

shanghaid:
a double-blind placebo controlled study shows that in disease A, treatment X reduces 1 year mortality by 50% with p less than 0.05.

isn't it the same to say, "there is a more than 95% chance that treatment X is effective," and "if treatment X really did not work, then there is a less than 5% chance that we would get same results on repeated testing?"


Don't confuse the treatment with the test.

"There is a more than 95% chance that our testing is accurate" is a better way of putting it, I think.

At least, that's what my brain worked it out to be. Separate the test from the process of testing.

Like, for instance, in the HIV vaccine study, what is the probability that the people who didn't get HIV were simply not exposed just right?

Where the hell is ABBW?

keytronic: But for medicine, I just can't see how anything bigger than .01 is tolerable. As 006007 says, assuming there is no relationship between the drug and infection rates, if you ran 100 trials you would find a correlation at least as big as yours 5 times. That seems high to me. Now, run three trials and get a p

That's why, as the article says, the FDA requires two studies with a p of 0.05 or less - basically a combined 0.0025.

I'd like to say THANK YOU to subby and all the commenters: would you believe the trouble I was having with my poli sci stats homework last night? All about this stuff too...

Perhaps too coincidental...

There is a 95% probability that this article won't go viral.

The only thing I know about p-value is that there's none of it in my ool, and I want to keep it that way.

subby tells us that the article is concise and precise, but doesn't tell us whether it's significant.

wildsnowllama: Mentat: squidzilla: Mentat: Statisticians aren't normal.

No, they aren't. They're average.

That's just mean.

It's a standard deviation...

The do tend to regress.

Seven frequent p-value misunderstandings (the following statements are true):

1. The p-value is not the probability that the null hypothesis is true.
2. The p-value is not the probability that a finding is "merely a fluke."
3. The p-value is not the probability of falsely rejecting the null hypothesis.
4. The p-value is not the probability that a replicating experiment would not yield the same conclusion.
5. 1 - (p-value) is not the probability of the alternative hypothesis being true (see (1)).
6. The significance level of the test is not determined by the p-value.
7. The p-value does not indicate the size or importance of the observed effect.

Twelve p-value misconceptions (the following statements are false):

1. If P = .05, the null hypothesis has only a 5% chance of being true.
2. A nonsignificant difference (eg, P ≥.05) means there is no difference between groups.
3. A statistically significant finding is clinically important.
4. Studies with P values on opposite sides of .05 are conflicting.
5. Studies with the same P value provide the same evidence against the null hypothesis.
6. P = .05 means that we have observed data that would occur only 5% of the time under the null hypothesis.
7. P = .05 and P ≤.05 mean the same thing.
8. P values are properly written as inequalities (eg, "P ≤.02" when P = .015)
9. P = .05 means that if you reject the null hypothesis, the probability of a type I error is only 5%.
10. With a P = .05 threshold for significance, the chance of a type I error will be 5%.
11. You should use a one-sided P value when you don't care about a result in one direction, or a difference in that direction is impossible.
12. A scientific conclusion or treatment policy should be based on whether or not the P value is significant.

There is a 95% probability that this article won't go viral.

Your P value is too high.

I tend to relate to the side article:

"A little fellatio goes a long way."

/A little? Mine is as big as a house.
//Obscure???

I'm currently in a graduate level business statistics class right now, and we're right in the midst of p-values...... and I'm still having trouble understanding this article.

megalynn44: I'm currently in a graduate level business statistics class right now, and we're right in the midst of p-values...... and I'm still having trouble understanding this article.

Your b-school sucks, then. :)

/MBA candidate myself
//graduate in about a month
///hopefully

I have a masters degree in biostatistics, and that article made me jizz myself.

I have an honours degree in statistics and 20 years professional experience working as one.

Ultimately, the expert's challenge is communicating what we know, clearly and concisely, to laypeople.

That article was far from concise. In fact it was decidedly unhelpful.

I eat mop: I have an honours degree in statistics and 20 years professional experience working as one.

So how much does working as an honors degree in stats pay?

/dont answer
//pedantic