Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?
DEAL: For $25 - Add A Second Phone Number To Your Smartphone for life! Use promo code SLASHDOT25. Also, Slashdot's Facebook page has a chat bot now. Message it for stories and more. Check out the new SourceForge HTML5 Internet speed test! ×

Comment Re:PET/MRI and statistics are poor bed partners (Score 1) 191

I agree that this is a very clear-cut point. The abstract simply does not say whether or not that p value is corrected. The article does. I refer you specifically to table 2, which provides the corrected p values for both of the ROIs. It lists them both as 0.05. I apologize if I'm reading this wrong, but I can't fathom what the basis could be for your disagreement, unless you were forced to guess without access to the full article.

If you're going to dispute my other two points, then I'm comfortable with that. There's a tidy little literature on the stupidity of retrospective power analysis. So far as I know, there's no opposing literature whatsoever. But of course academic statisticians are argumentative, so I'm sure you can find someone to disagree with just about anything, even if no one (to my knowledge) has been willing to make the point in a decent journal.

I'd also be shocked if you could get a knowledgeable statistician to sign on to the view that there's an accepted adequate number of observations for a t-test without any other details about the study. A simple power analysis seems like it would dispute that pretty readily (I won't insult anyone by posting numbers, at least not yet). But since you haven't made any argument, I can't really argue the point.

Comment Re:PET/MRI and statistics are poor bed partners (Score 1) 191

I think your third point misses the point. 47 participants probably is reasonable, but there is no such thing as an "accepted number for a t-test" without some prior estimate of the effect size you're looking for and the expected variability. All that said, there is absolutely no point in arguing about an observed effect on the basis of its prior power (aka retrospective power analysis). Once you have the results, the only thing that matters is whether or not the statistics were done correctly (and then, of course, what you're going to make of it, etc.). I know you know this, but it bears repeating.

Also, I believe p=0.004 is the uncorrected p value. Brain imaging studies generally pay a heavy price in terms of correction for multiple comparisons. In this case, the corrected p value was 0.05, which just meets the normal arbitrary standard. From my quick reading, the more impressive (statistically) result is the relationship between the electric field magnitude and the change in signal associated with on vs. off. That came in at 0.001, uncorrected.

Comment Re:PET/MRI and statistics are poor bed partners (Score 1) 191

Do you have some basis for saying that 47 participants is too few? If you think you do, then I urge you to do some more careful reading of your texts on power analysis. And I specifically urge you to read some of the many fine articles by statisticians on the evils of retrospective power analysis. The bottom line of all this is that once the study is done, and there are findings, it doesn't matter if it had too few subjects. Either the statistics are valid or they're not. In this case, they're... well, they're close enough to valid that I'll give it to them. No study is perfect, but it's at best misleading to describe a study as invalid because it doesn't have some arbitrary number of subjects.

If I'm doing you an injustice, and you have some legitimate reason to suspect that their methods have in inflated false positive rate, then please post more details, I'm always happy to learn new things.

I will agree, though, that imaging studies don't tell us nearly as much about human behavior as many people seem to think, although the worst offenders are not studies like this, but cognitive studies that make unsupported claims about what processes underly patterns of activation. And of course you're right that news outlets are likely to jump on this to make some kind of unsupported point about cancer.

Comment Re:Crappy Study or Crappy Reporting? (Score 1) 191

I've read the JAMA article. They report effects in the temporal pole, which is in the temporal lobe, but not that part that's associated with auditory processing. There are also frontal lobe effects, not too near the part of the frontal lobe that's directly involved in speech. It's also hard to argue that the effect is due to listening, because subjects didn't know when the phones were on or off, and for the results to have worked out, they'd have to have been listening more when the phones were on than off.

The confusing paragraph you cite is easy to explain. There's a difference between "overall" effects and local effects. They found effects localized to where the phone was, but not overall effects in the whole brain. This is just a sloppy way of pointing out that the effects they found were localized to where the cell phone was.

Comment Re:Unsure (Score 1) 191

Actually, both researchers and and the NIH are reasonably clever, for the most part (and they're the same people, by the way). The clueless people are idiots who have no scientific training or experience whatsoever and think they can offer appropriate scientific criticisms of a study they haven't read.

Comment Re:Unsure (Score 1) 191

Actually, they did use a reference condition with another device that didn't involve radio waves. It was a cell phone turned off. This is a much more appropriate control than what you suggest, and is a pretty good control for this study. Listening to recordings on headphones would be a very poorly chosen control condition, for obvious reasons.

Comment Re:Great book (Score 1) 583

Ignoring the sarcasm, I guess your point is that copyright has nothing to do with promoting great work, only crap. This is obviously ridiculous. Crap will always be in vastly greater supply, and copyright means we'll get a lot more of both. There would certainly be both brilliant work and utter crap with or without copyright. But since copyright creates a financial incentive, it probably promotes crap more than it promotes good work. But that's probably a really narrow difference. I could equally well argue that without copyright, only the independently wealthy would be able to devote their time to creating great works of art. Even with copyright, there is very little opportunity to earn a living as an artist.

Copyright law in the US is a pile of garbage that does much more harm than good to the arts. But the absence of copyright law would also be a pile of garbage that would do much more harm than good to the arts.

Comment Re:Short rant about e-books. (Score 1) 259

As a very conflicted Kindle owner, I couldn't agree more. I'm especially concerned that I might not be able to read the ebooks I buy today on the device I want to use tomorrow. If that device is a Nook, or a Sony reader, or some new software that's better than what Amazon provides, then I have to ditch my entire ebook library or maintain multiple devices. Truthfully, I'm hoping that the licensing terms will change during my lifetime, and I won't have to feel like a criminal for wanting to do useful things with my e-books. I'm not optimistic, but then again, I wasn't optimistic about the music industry giving up on DRM either. I do find the Kindle much more comfortable than reading books on paper, at least for books that are all text, and I've been willing to overlook these issues partly because I still consider it an early adopter technology.

Comment Re:I disapprove of Approval Voting (Score 1) 416

Approval voting is probably the best method that has a chance in hell of being adopted any time soon, and it addresses some of the most disturbing weaknesses of the current system. I'm happy with that. If you think you can get Schulze voting approved, go for it. But you're extremely misguided in advising people not to advocate approval voting, which is realistically attainable (Schulze voting is not) and a huge improvement, even if it's "broken." Remember, no method is perfect, so if we agree with you that we should not adopt a broken method, then we should not vote, period.

Comment very silly discussion (Score 1) 459

Facebook is no more evil this week than it was last. According to the article, apps can now request access to your private info. When an app requests your ifno, you can say yes or no. If you trust facebook to implement this as stated (one poster has already tried it and found it to be as described), then what's the big deal? Apps could also just ask you for your social security number, or your credit card numbers, this just makes it easier for users who were going to be agreeable. Who cares? And if you don't trust facebook, then you don't have private information on there anyway. So again, what's the big deal? This is a complete non-story that's been escalated into some kind of scare story by people who can't read.

Comment Re:why stop at addresses and phone numbers? (Score 1) 459

That's just stupid. Some people find fb useful and know how to use it safely. Frankly, anyone who falls victim to this is just a complete idiot. So yes, complete idiots should delete their accounts immediately (aso their email accounts and their phone accounts, and they should remove doors from their houses). Everyone else has nothing to worry about. I don't know why this is even news. Most people, like me, probably assumed that apps could request whatever they wanted. Apps could already just ask you for your phone number if they really wanted it.

Comment Re:A really nasty trick (Score 1) 765

If you're saying it's FUD that firefox users will heave to pay for a decoder, then please explain how we can legally obtain a free one.

Using OS libraries doesn't address the problem, it just means you're using a commercial OS, or a free OS with a commercial decoder installed. $.20 is indeed very reasonable, but unless they're legally compelled to maintain that price level through the 2028, I'm skeptical. Drug dealers offer an even better deal -- first one's free.

The fact that mozilla has enough cash to deal with this is irrelevant. Not everyone who might want to write software to support web standards has that kind of money. Right now, I can write a free, functional web browser if I want, and distribute it for free to whomever I like. If H.264 is adopted, I won't be able to. No one without massive capital will be able to, not for at least 17 years.

As an aside, I'm not arguing as if WebM is unencumbered. When that turns out to be a problem, I'll fault Google for failing to dodge patent potholes then. But for now, I'm willing to believe that if they say it will be BSD-licensed (or whatever), it will be.

Comment Re:A really nasty trick (Score 1) 765

Anything that slows or halts the transition to a standard that depends on commercialized patents is good for users. If MPEG LA has its way, everyone who uses firefox (for example) will have to buy a proprietary plug-in to use youtube. If Google has its way (which I don't like either, but for different reasons), we won't. We shouldn't allow web video to be held hostage for the next 18 years by large corporations, whether it's a single corporation or a coalition. If WebM turns out to have patent troubles, that's a separate issue -- you can't fault Google because other parties want to prevent them from making it available for free. I really don't think the phrase "open standard" has any useful meaning here. H.264 is not open in any way that helps anyone.

As an aside, I really don't care if it advances Google's agenda, unless you can describe how that agenda hurts me. My devices are going to become slightly less spiffy (not obsolete) about ten minutes faster? Who cares? Google wants to delay the transition away from flash to hurt Apple? That's fine with me -- protecting Apple is not a good reason to rush to adopt an expensive commercial product as a web standard.

Slashdot Top Deals

Artificial intelligence has the same relation to intelligence as artificial flowers have to flowers. -- David Parnas