There's a trade-off between sensitivity and specificity. If you increase the threshold for "significance", you reduce the power to discover a significant effect when it truly does exist.
And a major part of the problem with scientific studies is that they are already underpowered. According to conventional wisdom, ideally, scientists should strive for a power of about 80% (i.e., an 80% chance of detecting an effect if it truly exists), but very few studies actually achieve power of this level. In many fields, the power is less than 50% and sometimes much less.
Underpowered studies result in two major problems: 1) Most obviously, an underpowered study results in a greater number of FALSE NEGATIVES. You fail to find a true effect. You will either publish your incorrect result of no effect. (And why should we consider published false positives to be any worse than false negatives?) Alternatively, perhaps you don't publish your study because you couldn't reach significance. This exacerbates the "file-drawer effect" and also results in wasted research dollars because the results aren't published. 2) Somewhat counterintuitively, underpowered studies are often also more likely to result in FALSE POSITIVES. This is because, when your power to detect a true effect is low, and if you test a large number of effects that are unlikely to be null, most of the hypotheses that you say are "significantly" non-null will actually be false positives. We would say that the "false discovery rate" tends to be very high when the power is low.
Reducing the level of significance will do little to address these problems, and in some cases may even exacerbate the problem.
The key is *to move away from the binary concept of "significance" altogether*. It's obviously artificial to have an arbitrary numerical cutoff for "matters" vs. "doesn't matter", and this is not what Ronald Fisher intended when he popularized the p-value or developed the concept of "significance".
What we should be doing is measuring and reporting effect sizes along with their credible intervals. While using priors that are based on our real state of knowledge. In other words, we should be doing Bayesian statistics.
Mod parent down for being completely wrong. How the hell did this get modded up in the first place?
You need 2382 votes to clinch. Hillary won 2,205 pledged delegates; Bernie won 1,846. There were 712 superdelegate votes. 1846 + 712 is 2558.
News flash: 2558 is greater than 2382. Bernie could have won based on superdelegates.
Had the DNC not rigged the contest against Bernie from the start, and had superdelegates actually paid attention to the polls and realized that Bernie is far more electable than Hillary, then Trump would have lost last night.
2016 marks the end of Apple brand loyalty. We have quite clearly reached the point where the roadmap Steve Jobs laid out has ended, and now Cook and Ives are on their own, screwing things up as they go.
The outrage about today's keynote at AppleInsider is palpable. Among the common complaints are:
- These computers are overpriced and underwhelming. The price of the entry-level MacBook Pro was bumped up hundreds of dollars, and all they did was increase the price and remove ports from it. (The entry-level model only has two Thunderbolt ports (USB, etc. have been removed), and one of the ports has to be used for charging! What kind of "Pro" computer is that???) - The mind boggles that they removed the "esc" key from a supposedly "Pro" computer. - They removed the MagSafe connector, which is arguably one of the greatest features of Apple's laptops. - The only connections are Thunderbolt 3, meaning that you will need a dongle for ~anything~ you want to connect. Do you own an iOS device? Better hope you have a USB-C adapter for it. - Removal of the SD drive.
Apparently Apple has also been sending out emails to some of its customers asking if they use features such as the headphone jack on their laptop. (Because of course, they're going to remove it from there as well.)
Next time you feel like doing something brash, like removing a ubiquitous, standardized, and completely functional feature from your products, such as the headphone jack in order to promote wireless listening, why don't you go ahead and make the Bluetooth headphones available FIRST?
Or better yet, don't eliminate the jack at all. This isn't the 3.5" floppy in the age of CDs that we're talking about here. The headphone jack works well, it's standard, and it's on literally almost EVERYTHING that outputs sound.
And now Apple is set to announce tomorrow that its professional line of laptops will be moving completely to USB-C.
The linked article from Phil Plaitt in the summary is more than four years old. It has nothing to do with the current "double dynamo" claims by Zharkova.
And nowhere in my comment do I say that we shouldn't actually put cameras on cops. I merely say that it won't actually "solve" the problem. As the article itself points out.
It has nothing to do with "perfect being the enemy of the good".
Yes, if only the cop had a camera during the Michael Brown stop, then I suppose his killing would have looked more like this: https://www.youtube.com/watch?...
Don't forget, the cops in that case knew they were being filmed. Here's another case where cops disgracefully killed someone when they knew they were on film. He had a weapon, but was at such a distance that he posed no threat at all ( https://www.youtube.com/watch?... ) And another one: https://www.youtube.com/watch?...
Any time that cops are in a store, they know they're being filmed on security cameras. Here's another "heroic" action by the cops, committing what any sane human would consider to be murder while they know that they are being filmed: https://www.youtube.com/watch?...
Obviously this is all anecdotal and not "scientific" compared with the study in the summary, but it should be clear that this problem of police violence is not going to be completely solved until the cultures of "shoot first and ask questions later" and "protect each other" within law enforcement are changed.
We are not ready to accept apologies from anybody," says Luis Jaime Castillo, the vice minister for cultural heritage. "Let them apologize after they repair the damage.
First, the damage cannot be repaired. But second, Greenpeace has NOT issued a real apology. Their disgraceful excuse for an apology is here: http://www.greenpeace.org/inte...
The obvious missing element is an apology for defacing a UNESCO World Heritage Site. Instead, they offer mere apologies for how things LOOK, and the typical "I'm sorry if anyone was offended" not-pology. Peru should throw all of the activists in Prison, and when the Executive Director shows up in Lima, lock him up too.
Meanwhile, as others have pointed out, the image of the message doesn't even look real in the first place, and they could have gotten the exact same image from Photoshop. Here's the worthless Greenpeace image: http://www.iflscience.com/site...
Not going to happen. You need to publish during the data collection period in order to continue getting the funding you need for data collection.
Few replication attempts are doing exactly the same thing as the original paper, for good reason.
Right, but replication of the experiment is the EXACT reason that we're making the data available. If you want to use the data for something else, that's fine, but if it's data that the original author is still using, then you should contact them about it first.
A partial solution, I think, is for a group such as yours to pre-plan the data use already when collecting it. So you decide from start to publish a subset of that data early and publish papers based on that. Then publish another subset for further results and so on.
Again, this is not realistic in the overwhelming majority of cases. One of the benefits of long-term studies are the unexpected findings. Imagine that I've been collecting data on a population of lemmings over the last 20 years. It seems to me that the lemmings have been getting smaller since I first started capturing them, so one day I decide to regress body size on year and I discover that the lemmings have indeed been shrinking, and I can show that it is probably linked to changes in vegetation driven by climate change. I shouldn't have to give away my entire 20-year data set (which I had been collecting for a different purpose) for anybody to use for any purpose in order for me to get this one study out in a timely fashion.
Besides, many researchers are already dealing with data sets that are >50 years old, and your "plan to release the data before you start collecting the data" suggestion is moot for those people with inherited data sets.
But what we really need is for data to be fully citeable.
Getting your data cited is not NEARLY the same as publishing. Not even close. To get academic positions, pay increases, grants, etc., you need authorship. No one really cares about how often your paper or your data has been cited. That info isn't even on your CV or your grant applications, so no one will even have a rough idea unless it's a particularly preeminent paper.
This is bad news for ecologists and others with long-term data sets. Some of these data sets require decades of time and millions of dollars to produce, and the primary investigators want to use the data they've generated for multiple projects. Current data licensing for PLOS ONE (and--as far as I know-- all others who insist on complete data archiving) means that when you publish your data set, it is out there for anyone to use for free for any purpose that they wish; not just for verification of the paper in question. There are plenty of scientists out there who poach free online data sets and mine them for additional findings.
Requiring full accessibility of data makes many people reticent to publish in such a journal, because it means giving away the data they were planning on using for future publications. A scientist's publication list is linked not only to their job opportunities and their pay grade, but also to the funding that they can get for future grants. And of course those grants are linked to continuing the funding of the long-term project that produced the data in the first place.
What is needed is a new licensing model for published data that says "anyone is free to use these data to replicate the results of the current study, however it CANNOT be used as a basis for new analyses without written consent of the primary investigator of this paper or until [XX] years after publication." Journals would also need to agree that they would not accept any publications based on data that was used without consent.
It seems to me that this arrangement would satisfy the need to get data out into the public domain while respecting the scientists who produced it in the first place.
To give you an idea of how ridiculously overpriced WhatsApp is (and Facebook as well), here's a selection of major American companies with a market cap less than what Facebook paid for WhatsApp.
Retail: Macy’s Gap Bed Bath & Beyond Tiffany & Co. Ralph Lauren Staples Avon
Tech: LinkedIn Netflix Xerox Nvidia
Travel: Marriott International MGM Resorts Hertz Delta Air Lines United Airlines American Airlines Southwest Airlines
Food: Chipotle Hershey’s J. M. Smucker Campbell Soup Tyson Foods Dr Pepper Snapple Group Monster Beverage Molson Coors Brewing
Other: Harley-Davidson Mattel Whirlpool Western Union H&R Block McGraw-Hill News Corp The Carlyle Group
The author of this article, Matt Ridley, is a known climate change denialist and of course the Wall Street Journal is owned by Rupert Murdoch and therefore operates under the same umbrella as Fox News.
I have no reason to trust the right-wing's interpretation of the IPCC document before it is officially announced and I can check it for myself. Why don't you try WAITING for it to be released before you start spreading this very likely BS.
Slashdot Mirror
Industrial welders can make six-figures easily. (Work on an oil rig.) Truck drivers and heavy machinery operators also.
And PS, so can nurses (at least in much of Canada), with just a bit of overtime.
Are you that white-collar that you have no idea that blue-collar people can actually earn quite well?
A large number of results that are likely to be null, I mean.
There's a trade-off between sensitivity and specificity. If you increase the threshold for "significance", you reduce the power to discover a significant effect when it truly does exist.
And a major part of the problem with scientific studies is that they are already underpowered. According to conventional wisdom, ideally, scientists should strive for a power of about 80% (i.e., an 80% chance of detecting an effect if it truly exists), but very few studies actually achieve power of this level. In many fields, the power is less than 50% and sometimes much less.
Underpowered studies result in two major problems:
1) Most obviously, an underpowered study results in a greater number of FALSE NEGATIVES. You fail to find a true effect. You will either publish your incorrect result of no effect. (And why should we consider published false positives to be any worse than false negatives?) Alternatively, perhaps you don't publish your study because you couldn't reach significance. This exacerbates the "file-drawer effect" and also results in wasted research dollars because the results aren't published.
2) Somewhat counterintuitively, underpowered studies are often also more likely to result in FALSE POSITIVES. This is because, when your power to detect a true effect is low, and if you test a large number of effects that are unlikely to be null, most of the hypotheses that you say are "significantly" non-null will actually be false positives. We would say that the "false discovery rate" tends to be very high when the power is low.
Reducing the level of significance will do little to address these problems, and in some cases may even exacerbate the problem.
The key is *to move away from the binary concept of "significance" altogether*. It's obviously artificial to have an arbitrary numerical cutoff for "matters" vs. "doesn't matter", and this is not what Ronald Fisher intended when he popularized the p-value or developed the concept of "significance".
What we should be doing is measuring and reporting effect sizes along with their credible intervals. While using priors that are based on our real state of knowledge. In other words, we should be doing Bayesian statistics.
This "grenade" story is not at all true.
Her arm was blown off by a propane-tank IED built by the protesters. The remains of the tanks were confiscated after the explosion, and her clothing was taken for testing.
http://www.nytimes.com/2016/11...
http://bearingarms.com/bob-o/2...
Mod parent down for being completely wrong. How the hell did this get modded up in the first place?
You need 2382 votes to clinch. Hillary won 2,205 pledged delegates; Bernie won 1,846. There were 712 superdelegate votes. 1846 + 712 is 2558.
News flash:
2558 is greater than 2382. Bernie could have won based on superdelegates.
Had the DNC not rigged the contest against Bernie from the start, and had superdelegates actually paid attention to the polls and realized that Bernie is far more electable than Hillary, then Trump would have lost last night.
2016 marks the end of Apple brand loyalty. We have quite clearly reached the point where the roadmap Steve Jobs laid out has ended, and now Cook and Ives are on their own, screwing things up as they go.
The outrage about today's keynote at AppleInsider is palpable. Among the common complaints are:
- These computers are overpriced and underwhelming. The price of the entry-level MacBook Pro was bumped up hundreds of dollars, and all they did was increase the price and remove ports from it. (The entry-level model only has two Thunderbolt ports (USB, etc. have been removed), and one of the ports has to be used for charging! What kind of "Pro" computer is that???)
- The mind boggles that they removed the "esc" key from a supposedly "Pro" computer.
- They removed the MagSafe connector, which is arguably one of the greatest features of Apple's laptops.
- The only connections are Thunderbolt 3, meaning that you will need a dongle for ~anything~ you want to connect. Do you own an iOS device? Better hope you have a USB-C adapter for it.
- Removal of the SD drive.
Apparently Apple has also been sending out emails to some of its customers asking if they use features such as the headphone jack on their laptop. (Because of course, they're going to remove it from there as well.)
This company has lost its mind.
http://www.bbc.com/news/techno...
Next time you feel like doing something brash, like removing a ubiquitous, standardized, and completely functional feature from your products, such as the headphone jack in order to promote wireless listening, why don't you go ahead and make the Bluetooth headphones available FIRST?
Or better yet, don't eliminate the jack at all. This isn't the 3.5" floppy in the age of CDs that we're talking about here. The headphone jack works well, it's standard, and it's on literally almost EVERYTHING that outputs sound.
And now Apple is set to announce tomorrow that its professional line of laptops will be moving completely to USB-C.
This company has lost its way and its mind.
It's called a "shadow CV". Haushofer is hardly the first to post one.
For example, from 2012:
https://dynamicecology.wordpre...
At first, I thought that said "Internet Fapping Glitch Turned a Random Kansas Farm Into a Digital Hell".
That would have been a different kind of story.
Unfamous enough that even the submitter didn't know how to spell his name. "Mims", not "Mimms". Kinda undermines your point.
Why is this news?
Related Onion video:
https://www.youtube.com/watch?...
The linked article from Phil Plaitt in the summary is more than four years old. It has nothing to do with the current "double dynamo" claims by Zharkova.
And nowhere in my comment do I say that we shouldn't actually put cameras on cops. I merely say that it won't actually "solve" the problem. As the article itself points out.
It has nothing to do with "perfect being the enemy of the good".
Yes, if only the cop had a camera during the Michael Brown stop, then I suppose his killing would have looked more like this: https://www.youtube.com/watch?...
Don't forget, the cops in that case knew they were being filmed. Here's another case where cops disgracefully killed someone when they knew they were on film. He had a weapon, but was at such a distance that he posed no threat at all ( https://www.youtube.com/watch?... ) And another one: https://www.youtube.com/watch?...
Any time that cops are in a store, they know they're being filmed on security cameras. Here's another "heroic" action by the cops, committing what any sane human would consider to be murder while they know that they are being filmed: https://www.youtube.com/watch?...
And of course don't expect some of the footage not to go "missing" ( http://www.washingtonpost.com/... ), and don't expect the footage to even be released ( http://www.citylab.com/crime/2... ). And even if it goes to a grand jury, don't expect the District Attorney not to knowingly put a liar on the stand and throw the case: https://www.youtube.com/watch?...
Obviously this is all anecdotal and not "scientific" compared with the study in the summary, but it should be clear that this problem of police violence is not going to be completely solved until the cultures of "shoot first and ask questions later" and "protect each other" within law enforcement are changed.
We are not ready to accept apologies from anybody," says Luis Jaime Castillo, the vice minister for cultural heritage. "Let them apologize after they repair the damage.
First, the damage cannot be repaired. But second, Greenpeace has NOT issued a real apology. Their disgraceful excuse for an apology is here:
http://www.greenpeace.org/inte...
The obvious missing element is an apology for defacing a UNESCO World Heritage Site. Instead, they offer mere apologies for how things LOOK, and the typical "I'm sorry if anyone was offended" not-pology. Peru should throw all of the activists in Prison, and when the Executive Director shows up in Lima, lock him up too.
Meanwhile, as others have pointed out, the image of the message doesn't even look real in the first place, and they could have gotten the exact same image from Photoshop. Here's the worthless Greenpeace image:
http://www.iflscience.com/site...
And here's the damage the fuckers caused:
http://cdn.zmescience.com/wp-c...
Prison sentences for all.
What kind of shirt were they wearing when they made the announcement? The world needs to know.
Release all the papers when you release the data.
Not going to happen. You need to publish during the data collection period in order to continue getting the funding you need for data collection.
Few replication attempts are doing exactly the same thing as the original paper, for good reason.
Right, but replication of the experiment is the EXACT reason that we're making the data available. If you want to use the data for something else, that's fine, but if it's data that the original author is still using, then you should contact them about it first.
A partial solution, I think, is for a group such as yours to pre-plan the data use already when collecting it. So you decide from start to publish a subset of that data early and publish papers based on that. Then publish another subset for further results and so on.
Again, this is not realistic in the overwhelming majority of cases. One of the benefits of long-term studies are the unexpected findings. Imagine that I've been collecting data on a population of lemmings over the last 20 years. It seems to me that the lemmings have been getting smaller since I first started capturing them, so one day I decide to regress body size on year and I discover that the lemmings have indeed been shrinking, and I can show that it is probably linked to changes in vegetation driven by climate change. I shouldn't have to give away my entire 20-year data set (which I had been collecting for a different purpose) for anybody to use for any purpose in order for me to get this one study out in a timely fashion.
Besides, many researchers are already dealing with data sets that are >50 years old, and your "plan to release the data before you start collecting the data" suggestion is moot for those people with inherited data sets.
But what we really need is for data to be fully citeable.
Getting your data cited is not NEARLY the same as publishing. Not even close. To get academic positions, pay increases, grants, etc., you need authorship. No one really cares about how often your paper or your data has been cited. That info isn't even on your CV or your grant applications, so no one will even have a rough idea unless it's a particularly preeminent paper.
This is bad news for ecologists and others with long-term data sets. Some of these data sets require decades of time and millions of dollars to produce, and the primary investigators want to use the data they've generated for multiple projects. Current data licensing for PLOS ONE (and--as far as I know-- all others who insist on complete data archiving) means that when you publish your data set, it is out there for anyone to use for free for any purpose that they wish; not just for verification of the paper in question. There are plenty of scientists out there who poach free online data sets and mine them for additional findings.
Requiring full accessibility of data makes many people reticent to publish in such a journal, because it means giving away the data they were planning on using for future publications. A scientist's publication list is linked not only to their job opportunities and their pay grade, but also to the funding that they can get for future grants. And of course those grants are linked to continuing the funding of the long-term project that produced the data in the first place.
What is needed is a new licensing model for published data that says "anyone is free to use these data to replicate the results of the current study, however it CANNOT be used as a basis for new analyses without written consent of the primary investigator of this paper or until [XX] years after publication." Journals would also need to agree that they would not accept any publications based on data that was used without consent.
It seems to me that this arrangement would satisfy the need to get data out into the public domain while respecting the scientists who produced it in the first place.
To give you an idea of how ridiculously overpriced WhatsApp is (and Facebook as well), here's a selection of major American companies with a market cap less than what Facebook paid for WhatsApp.
Retail:
Macy’s
Gap
Bed Bath & Beyond
Tiffany & Co.
Ralph Lauren
Staples
Avon
Tech:
LinkedIn
Netflix
Xerox
Nvidia
Travel:
Marriott International
MGM Resorts
Hertz
Delta Air Lines
United Airlines
American Airlines
Southwest Airlines
Food:
Chipotle
Hershey’s
J. M. Smucker
Campbell Soup
Tyson Foods
Dr Pepper Snapple Group
Monster Beverage
Molson Coors Brewing
Other:
Harley-Davidson
Mattel
Whirlpool
Western Union
H&R Block
McGraw-Hill
News Corp
The Carlyle Group
The author of this article, Matt Ridley, is a known climate change denialist and of course the Wall Street Journal is owned by Rupert Murdoch and therefore operates under the same umbrella as Fox News.
Supposed leaks from the IPCC document have already been mischaracterized in the right-wing media. See, for example, Phil Plait's demolition of them here:
http://www.slate.com/blogs/bad_astronomy/2013/09/10/climate_change_sea_ice_global_cooling_and_other_nonsense.html
Or if you prefer your demolition in video format:
http://www.youtube.com/watch?v=lH5D9P6KYfY
I have no reason to trust the right-wing's interpretation of the IPCC document before it is officially announced and I can check it for myself. Why don't you try WAITING for it to be released before you start spreading this very likely BS.
Since OP forgot it, you can learn more about Khan Academy here: http://khaaan.com/
Share fluids, not passwords.
Not many people know it, but the apex of the Washington Monument is made of aluminum. At the time, it was the largest piece ever crafted anywhere in the world and it was a precious metal. Only two years later, aluminum became completely worthless when the Hallâ"Héroult process for mass production of pure aluminum was discovered.
Eat shit, faggoot.