Suppose 1 out of 10,000 people in the USA have HIV. Suppose an HIV test has a 99.9% chance of detecting HIV in infected people, and a 99.9% chance of testing negative in people without HIV.
Then, suppose we test 10,000 people.
Among the infected, we will get an average of.999 people who test positive. Among the non-infected, we will get.001 * 9999 = 9.99 people who test positive.
Thus, among people who tested positive, only.999/9.99 = about 10% will actually be infected.
I'm in grad school in a health-related field. I've had more than one professor tell me that the ink blot tests are no longer taken seriously because they are too subjective.
I'm a statistician working for a health insurance company. If there is redundant testing, we are the first to know about it, and we will not pay for it. The health insurance industry has had no problems implementing IT technology; we have very good databases. Health care providers have traditionally been in charge of treating patients rather than keeping records, so we have served as a default IT infrastructure. One of the main obstacles to implementing IT in health care environments is the lack of computer literacy. A very large number of nurses and physicians still have no idea how to read e-mail or surf the web, even in 2009; until a few months ago, I used to work in a health care environment and witnessed it firsthand. There is also a shortage of expertise in IT professionals who understand the complexities of health care; a typical computer science graduate knows nothing about medicine and is hung up on the "healing power of echinacea" or whatever. Both IT professionals and health care professionals regard themselves as smart people, they do not like to look stupid, so they resist learning skills unfamiliar to them. In many markets, health care providers are given financial incentives to submit their claims electronically. Larger hospitals and their affiliated clinics can afford to implement such measures, but private practices often cannot. In many patients, there are extenuating circumstances that require the patient to receive treatments that deviate from standard procedure, so judgements still need to be made by humans on a case by case basis; it is not as simple as issuing tickets to airline passengers or shuffling boxes around in a warehouse. In summary, there are good structural reasons why the health care industry has been more resistant to implementing IT, not just "greed" or "conspiracies".
Let's compare apples to apples. What if you were in a rocket engine testing facility where dozens of rocket engines are simultaneously in danger of failing, and all the rocket engines had to keep running at all costs? Would you want to go through the formality of checklists? If ICU patient #37 died while doctors were going through the checklist for patient #14, won't the media headlines scream that "ICU patients die while doctors are held up by hospital bureacracy"?
There is an interesting recent article that shows how fire extinguishers are harmful to public health; 4% of fire deaths each year are caused by amateur operators who attempt to extinguish flames with their "mandatory" fire extinguishers. As the paper states, "As described above, in terms of life safety (as opposed to preventing material losses), attempting to extinguish a fire is not effective. For a fire extinguisher to work, a fire must be small, accessible, and not producing large quantities of smoke. In the face of such a fire, the safest course of action is to evacuate the area.... Fire extinguishers require extensive training to be used effectively. Using a fire extinguisher exposes the operator to physical danger and in terms of life safety--evacuation, as opposed to attempting to extinguish the fire probably represents the best option."
Mass hysteria has mandated "safety" laws before, which have turned out to do more harm than good in the long run.
So the checklist reduced infection rates. But then, how many lives were lost for other reasons because medical staff wasted time on these checklists, while nurses enjoyed obstructing physicians?
A few years ago, people were absolutely sure that millions of lives could be saved by having AEDs everywhere. How many lives are being lost by amateur do-gooders who unnecessarily electrocute people who are not having heart attacks? Plenty of medical problems can appear like heart attacks to the untrained eye.
Many schools that are not too famous have very good departments in particular fields. The college guides focus almost exclusively on the undergraduate experience, but from my grad student point of view, undergrads are just so much background noise.
Probabilities of all possible outcomes will add up to 1, but odds are p/(1-p), where p = probability of a given event. Odds can vary between 0 and infinity.
Logistic regression predicts the log-odds of a given event (which can be exponentiated to predict odds, or converted to a probability.)
Spread all around the economy, ranging from Google to mom'n'pop shops. The linux billionaires are those who _use_ linux and save money. Coincidentally, the very same are often those who invest time back to solve their own problems, as the money they save far, far outweighs the money it'd cost to roll their own from scratch.
Do they really save billions of dollars? Does every company have to become a Linux specialist to "invest time back to solve their own problems"? Or were you going to argue that "that's what vendors like Red Hat are for", in which case it begins to look just like a for-profit proprietary operating system anyway?
The fact that someone is making money from monopoly protections does not mean that it's good for the economy as a whole. We could hand out monopoly rights for air, and you'd get a huge AirCo, developing amazing technology for measuring how much air each person was using and charging for it. They'd certainly make money, but we'd all be poorer by paying for a resources that would have been produced anyway.
Ok, so if Red Hat makes money from monopoly protection by supporting only its brand of Linux, what is it then? Does everyone in the world have to become an expert in solving complex system engineering problems to force software to be truly free? Or does such a scheme incur heavy costs?
Linux, BSD, and all Free Software proves that software would be produced anyway.
There is plenty of "free" software out there, but which one of them are trustworthy? Which one of them have funny viruses or other malware snuck in?
If anyone could just copy chips we'd get the same economy there. There are many 'open chip' projects around.
So where are the open chip billionaires?
The purpose of the economy isn't about 'making money'.
Really? Last I checked, the purpose of most people's lives is to pay the bills. The only people I know who think otherwise are decadent techies who work at a for-profit software firm and write their funny freeware as a hobby, and maybe take payments from malware writers on the side.
The purpose of the free market economy is to maximize the creation of wealth by encouraging competition in overcharging sectors and constantly lowering the costs of production.
What happens when "lowering the costs of production" imposes its own costs through lowered ethics, societal and environmental harm?
When the cost of production reaches zero we've all won; we've got infinite wealth.
The cost of producing spam is essentially zero, and spam generates wealth. Have we "all won"? Have we created infinite wealth?
Silicon Valley has and continues to derive the vast majority of its income from intellectual property protections for its software. I pointed this out on Techdirt, so the commenters there hemmed and hawed with their red herring arguments about how Microsoft does not make money from software written 14 years ago. Regardless, Microsoft (which is no longer a Silicon Valley firm, I know) would make no money today if XP and Vista were free. Intel would make no money if anybody could just copy Intel chips. If they were free, nobody would bother with Linux. Where are the linux billionaires? Nor would biotech companies make any money if anybody could just copy their inventions. Sun, AIX, etc. all made fortunes in their time from selling proprietary flavors of Unix. SAS and SPSS are the industry standards for statistical computing, and they are proprietary, intellectually protected, for-profit firms.
Japanese electronics companies used to have cargo ships with mainframes which performed data processing tasks in international waters near San Francisco. (Back when Japanese labor costs were low, and computers were rare enough that there was benefit to having a mobile computer.)
#2 and #4 are having it both ways. If the Aryan race goes extinct, then the beliefs of present-day French Aryans will have no more relevance to the future residents of Frankistan than the beliefs of the Anasazi or Yavapai people have relevance to Americans today. The present-day French will just be a minor footnote in history textbooks: "Ancient Frankistan was once populated by a violent race of cannibal Gauls, who later converted to Christianity, but lost their Christian identity and went extinct".
6) What's wrong with Buddhists? Buddhism has easily got to be the single most inoffensive religion on the planet.
A look at the history of Buddhist countries will reveal them to be every bit as violent as European or Middle Eastern countries.
Will it become a place where Latin American and African Christians live in tense coexistence with Moslems, while absentee landlords from Asia own everything? Will whites become so rare that New Yorkers will stare in fascination at white people? Jews will have long since have converted to Buddhism or intermarried with others, that they are a regarded as a mysterious ancient people like the Druids or Manicheans. The world's economic center of gravity will have long since shifted South and East, so New York will be a historical curiosity like Philadelphia or Pittsburgh today. (In their time, Philadelphia and Pittsburgh were the apex of American culture and technology.) At the request of France's Islamic government, the Statue of Liberty will be replaced with a statue of Sayyid Qutb, every schoolkid will take museum trips to the "Palestinian Holocaust Museum", Chinese financiers will turn Central Park into a replica of the Forbidden City, while trendy New Yorkers will receive cosmetic gene therapy to look more Arab, African, or Hispanic.
One of these things isn't like the others!;) How in the bloody hell can you compare spewing racist rhetoric and posting medical records?
Simple. What if people post realistic-looking records of you, making you appear as an international sex predator or other criminal? What if this has real consequences, such as this information gets sent to your work supervisor, to all your friends, to your university?
And why should any of that be illegal? And if people want to post their financial or medical records they can go ahead and do that- but they should be prepared for the consequences, especially with that medical record data...
If somebody posts your financial and medical records online, do you think they can "go ahead and do that"? Should you be prepared for the consequences and just accept your fate that others can do whatever they want to you?
This is a way to make up for the deficiencies of Japan's legal system. Under the present system, people can post anonymously online , often through the "2ch" bulletin board, to make up false accusations about others, post their financial and medical records online, their bank account numbers, spew racist rhetoric, make death threats, etc. Japanese courts have shown no interest in enforcing the egregious violations of other people's rights. At present, there is a whole subculture of professional losers, the "NEETs" in their 20s and 30s who live at home with their parents and don't work, who spend their lives posting this stuff on the web.
If 2 guys in a garage can invent a cure for AIDS, then I would like to see that.
In the past, plenty of brilliant physicists, businessmen, etc. came into medical research thinking they can teach doctors a lesson. Their elegant theories held no water in the health science context, so they walked away embarrassed. Andy Grove can believe in his "herbal medicine cures" and take all the cheapshots he likes.
In Japan, it's often more difficult to get a refund for a defective product. Also, consumer safety standards tend to be lower. Things that are illegal in the USA are legal in Japan, such as a batting cage that serves hardballs at 200km/h:
I majored in liberal arts in college because I didn't know what I wanted to do. But I became interested in science, so I overcame a lot of prejudice to get into a statistics graduate program. I had many professors tell me outright that I don't have a science background, therefore I'm incapable of getting into statistics. I've completed my M.S. at a major university and I'm working on my PhD now while working full time.
Statistically speaking, you don't need 10 subjects per covariate. Where did you pull that number from?
If you want other statisticians to take you seriously, you do. I learned it from my professors.
Let's say we have 100 subjects (50 cases, 50 controls), and 100 tests that we're performing. Let's say one of those tests segregates perfectly with our subjects. Bam! Huge result statistically.
Or maybe poor experimental design? Were the cases truly identical to the controls? In a clinical trial, how do you find 50 cases and 50 controls who are exactly alike? How do you ensure that they really underwent identical treatment, as opposed to overly eager doctors and nurses who influenced the outcome? "Double blind" trials, in the real world, are a sham -- doctors often find out who the controls and treatment patients are anyway, based on "safety" excuses.
The probability of that result is astonishingly small (p ~ 10^-30). That means unless we are performing close to 10^30 tests, we can believe this result.
Or did the researchers exclude data they didn't like, to make the p-value small? Are we sure that genetics researchers are not under huge pressure to come up with significant results, given their scarce grants?
First, negative results are in fact published in this field.
Very rare. How often do we see such articles on, say, Science magazine? Scientific American? Slashdot? Or even the Journal of Genetic Epidemiology?
See here for an example about this very study!
How sure are they that the locus in question really came from 18q22? One of the dirty secrets of this field is that people are less sure about which chromosome the DNA came from than they care to admit.
or here for another diabetes replication.
So they passed data through two cohorts, and found several loci, "emphasizing the contribution of multiple variants of modest effect" -- a euphemism for "noise".
All the single-gene disorders with high penetrance have been known since a long time ago. Discovering the genetic origin of diseases such as hypertension or diabetes with (assumed) polygenic origins, which include (assumed) environmental interactions, remain a fantasy that has been "just a few years away" for a long time. In all likelihood, many different mechanisms underlie each disease. While some forms of mental retardation such as Trisomy-21 are easy to test for, there are dozens of other forms of mental retardation with uncertain genetic origins, possibly not genetic at all. Of course, there are other diseases with known genetic mechanisms, such as Huntington's Disease, which is unlikely to occur until old age -- should babies be screened for that? Should kids be told to expect short lives, after which they will become drooling basket cases?
Speaking as a statistical genetics insider, I can tell you that the replication of results in this field is very poor. A team of scientists somewhere will announce they found a gene for XYZ, which is reported in scientific journals and mainstream media -- however, the findings fail to be replicated by other scientists, and the negative results are usually not published. Over the years, hundreds of scientists have claimed to find genes responsible for diabetes, hypertension, autism, etc.
Since there are tens of thousands of genes in the genome, a study with 17,000 subjects makes for less than one subject per gene. (Exactly how many "genes" are in the genome anyway? What exactly defines a "gene"? That is another vast topic.)
Statistically speaking, there should be at least ten subjects per covariate (gene) tested. There is a great deal of hoo-haa over microarrays, but the more you learn about microarrays, the more you will learn just how unreliable they are. The same "disease" can have vastly different pathophysiologies and genetic origins across population groups. Epigenetics, penetrance, expressivity, intron effects -- all multiply the complexity exponentially.
In short, genetics is to biology what nuclear fusion is to physics -- a promising technology that will remain a "few years away" for decades to come.
For decades, medical scientists were sure that obesity raises cancer risk. However, a slew of large recent studies have cast this theory into serious doubt -- obesity has little, if any, statistical association with many cancers. I would take this study with a big grain of salt.
Slashdot Mirror
Suppose 1 out of 10,000 people in the USA have HIV. Suppose an HIV test has a 99.9% chance of detecting HIV in infected people, and a 99.9% chance of testing negative in people without HIV.
Then, suppose we test 10,000 people.
Among the infected, we will get an average of .999 people who test positive. Among the non-infected, we will get .001 * 9999 = 9.99 people who test positive.
Thus, among people who tested positive, only .999/9.99 = about 10% will actually be infected.
I'm in grad school in a health-related field. I've had more than one professor tell me that the ink blot tests are no longer taken seriously because they are too subjective.
I'm a statistician working for a health insurance company. If there is redundant testing, we are the first to know about it, and we will not pay for it. The health insurance industry has had no problems implementing IT technology; we have very good databases. Health care providers have traditionally been in charge of treating patients rather than keeping records, so we have served as a default IT infrastructure. One of the main obstacles to implementing IT in health care environments is the lack of computer literacy. A very large number of nurses and physicians still have no idea how to read e-mail or surf the web, even in 2009; until a few months ago, I used to work in a health care environment and witnessed it firsthand. There is also a shortage of expertise in IT professionals who understand the complexities of health care; a typical computer science graduate knows nothing about medicine and is hung up on the "healing power of echinacea" or whatever. Both IT professionals and health care professionals regard themselves as smart people, they do not like to look stupid, so they resist learning skills unfamiliar to them. In many markets, health care providers are given financial incentives to submit their claims electronically. Larger hospitals and their affiliated clinics can afford to implement such measures, but private practices often cannot. In many patients, there are extenuating circumstances that require the patient to receive treatments that deviate from standard procedure, so judgements still need to be made by humans on a case by case basis; it is not as simple as issuing tickets to airline passengers or shuffling boxes around in a warehouse. In summary, there are good structural reasons why the health care industry has been more resistant to implementing IT, not just "greed" or "conspiracies".
Let's compare apples to apples. What if you were in a rocket engine testing facility where dozens of rocket engines are simultaneously in danger of failing, and all the rocket engines had to keep running at all costs? Would you want to go through the formality of checklists? If ICU patient #37 died while doctors were going through the checklist for patient #14, won't the media headlines scream that "ICU patients die while doctors are held up by hospital bureacracy"?
There is an interesting recent article that shows how fire extinguishers are harmful to public health; 4% of fire deaths each year are caused by amateur operators who attempt to extinguish flames with their "mandatory" fire extinguishers. As the paper states, "As described above, in terms of life safety (as opposed to preventing material losses), attempting to extinguish a fire is not effective. For a fire extinguisher to work, a fire must be small, accessible, and not producing large quantities of smoke. In the face of such a fire, the safest course of action is to evacuate the area.... Fire extinguishers require extensive training to be used effectively. Using a fire extinguisher exposes the operator to physical danger and in terms of life safety--evacuation, as opposed to attempting to extinguish the fire probably represents the best option."
Mass hysteria has mandated "safety" laws before, which have turned out to do more harm than good in the long run.
Prog Cardiovasc Dis. 2008 Nov-Dec;51(3):204-12.
So the checklist reduced infection rates. But then, how many lives were lost for other reasons because medical staff wasted time on these checklists, while nurses enjoyed obstructing physicians?
A few years ago, people were absolutely sure that millions of lives could be saved by having AEDs everywhere. How many lives are being lost by amateur do-gooders who unnecessarily electrocute people who are not having heart attacks? Plenty of medical problems can appear like heart attacks to the untrained eye.
Many schools that are not too famous have very good departments in particular fields. The college guides focus almost exclusively on the undergraduate experience, but from my grad student point of view, undergrads are just so much background noise.
Are you mixing up probability and odds?
Probabilities of all possible outcomes will add up to 1, but odds are p/(1-p), where p = probability of a given event. Odds can vary between 0 and infinity.
Logistic regression predicts the log-odds of a given event (which can be exponentiated to predict odds, or converted to a probability.)
Spread all around the economy, ranging from Google to mom'n'pop shops. The linux billionaires are those who _use_ linux and save money. Coincidentally, the very same are often those who invest time back to solve their own problems, as the money they save far, far outweighs the money it'd cost to roll their own from scratch.
Do they really save billions of dollars? Does every company have to become a Linux specialist to "invest time back to solve their own problems"? Or were you going to argue that "that's what vendors like Red Hat are for", in which case it begins to look just like a for-profit proprietary operating system anyway?
The fact that someone is making money from monopoly protections does not mean that it's good for the economy as a whole. We could hand out monopoly rights for air, and you'd get a huge AirCo, developing amazing technology for measuring how much air each person was using and charging for it. They'd certainly make money, but we'd all be poorer by paying for a resources that would have been produced anyway.
Ok, so if Red Hat makes money from monopoly protection by supporting only its brand of Linux, what is it then? Does everyone in the world have to become an expert in solving complex system engineering problems to force software to be truly free? Or does such a scheme incur heavy costs?
Linux, BSD, and all Free Software proves that software would be produced anyway.
There is plenty of "free" software out there, but which one of them are trustworthy? Which one of them have funny viruses or other malware snuck in?
If anyone could just copy chips we'd get the same economy there. There are many 'open chip' projects around.
So where are the open chip billionaires?
The purpose of the economy isn't about 'making money'.
Really? Last I checked, the purpose of most people's lives is to pay the bills. The only people I know who think otherwise are decadent techies who work at a for-profit software firm and write their funny freeware as a hobby, and maybe take payments from malware writers on the side.
The purpose of the free market economy is to maximize the creation of wealth by encouraging competition in overcharging sectors and constantly lowering the costs of production.
What happens when "lowering the costs of production" imposes its own costs through lowered ethics, societal and environmental harm?
When the cost of production reaches zero we've all won; we've got infinite wealth.
The cost of producing spam is essentially zero, and spam generates wealth. Have we "all won"? Have we created infinite wealth?
Silicon Valley has and continues to derive the vast majority of its income from intellectual property protections for its software. I pointed this out on Techdirt, so the commenters there hemmed and hawed with their red herring arguments about how Microsoft does not make money from software written 14 years ago. Regardless, Microsoft (which is no longer a Silicon Valley firm, I know) would make no money today if XP and Vista were free. Intel would make no money if anybody could just copy Intel chips. If they were free, nobody would bother with Linux. Where are the linux billionaires? Nor would biotech companies make any money if anybody could just copy their inventions. Sun, AIX, etc. all made fortunes in their time from selling proprietary flavors of Unix. SAS and SPSS are the industry standards for statistical computing, and they are proprietary, intellectually protected, for-profit firms.
I remember a line from that movie where the mentor guy tells the kid not to end up like the chess-playing bums in the park.
They left out the part where the kid turns into one of them.
Japanese electronics companies used to have cargo ships with mainframes which performed data processing tasks in international waters near San Francisco. (Back when Japanese labor costs were low, and computers were rare enough that there was benefit to having a mobile computer.)
#2 and #4 are having it both ways. If the Aryan race goes extinct, then the beliefs of present-day French Aryans will have no more relevance to the future residents of Frankistan than the beliefs of the Anasazi or Yavapai people have relevance to Americans today. The present-day French will just be a minor footnote in history textbooks: "Ancient Frankistan was once populated by a violent race of cannibal Gauls, who later converted to Christianity, but lost their Christian identity and went extinct".
6) What's wrong with Buddhists? Buddhism has easily got to be the single most inoffensive religion on the planet.
A look at the history of Buddhist countries will reveal them to be every bit as violent as European or Middle Eastern countries.
p.s. I am not white.
Will it become a place where Latin American and African Christians live in tense coexistence with Moslems, while absentee landlords from Asia own everything? Will whites become so rare that New Yorkers will stare in fascination at white people? Jews will have long since have converted to Buddhism or intermarried with others, that they are a regarded as a mysterious ancient people like the Druids or Manicheans. The world's economic center of gravity will have long since shifted South and East, so New York will be a historical curiosity like Philadelphia or Pittsburgh today. (In their time, Philadelphia and Pittsburgh were the apex of American culture and technology.) At the request of France's Islamic government, the Statue of Liberty will be replaced with a statue of Sayyid Qutb, every schoolkid will take museum trips to the "Palestinian Holocaust Museum", Chinese financiers will turn Central Park into a replica of the Forbidden City, while trendy New Yorkers will receive cosmetic gene therapy to look more Arab, African, or Hispanic.
One of these things isn't like the others! ;) How in the bloody hell can you compare spewing racist rhetoric and posting medical records?
Simple. What if people post realistic-looking records of you, making you appear as an international sex predator or other criminal?
What if this has real consequences, such as this information gets sent to your work supervisor, to all your friends, to your university?
And why should any of that be illegal? And if people want to post their financial or medical records they can go ahead and do that- but they should be prepared for the consequences, especially with that medical record data...
If somebody posts your financial and medical records online, do you think they can "go ahead and do that"? Should you be prepared for the consequences and just accept your fate that others can do whatever they want to you?
This is a way to make up for the deficiencies of Japan's legal system. Under the present system, people can post anonymously online , often through the "2ch" bulletin board, to make up false accusations about others, post their financial and medical records online, their bank account numbers, spew racist rhetoric, make death threats, etc. Japanese courts have shown no interest in enforcing the egregious violations of other people's rights. At present, there is a whole subculture of professional losers, the "NEETs" in their 20s and 30s who live at home with their parents and don't work, who spend their lives posting this stuff on the web.
I remember reading the same story in the early 1990s.
If 2 guys in a garage can invent a cure for AIDS, then I would like to see that.
In the past, plenty of brilliant physicists, businessmen, etc. came into medical research thinking they can teach doctors a lesson. Their elegant theories held no water in the health science context, so they walked away embarrassed. Andy Grove can believe in his "herbal medicine cures" and take all the cheapshots he likes.
In Japan, it's often more difficult to get a refund for a defective product. Also, consumer safety standards tend to be lower. Things that are illegal in the USA are legal in Japan, such as a batting cage that serves hardballs at 200km/h:
http://jp.youtube.com/watch?v=NMSgAs7VLTI
I remember seeing articles in 1997 hyping up transdermal microneedle skin patches. Not much has come of it since.
I majored in liberal arts in college because I didn't know what I wanted to do. But I became interested in science, so I overcame a lot of prejudice to get into a statistics graduate program. I had many professors tell me outright that I don't have a science background, therefore I'm incapable of getting into statistics. I've completed my M.S. at a major university and I'm working on my PhD now while working full time.
Statistically speaking, you don't need 10 subjects per covariate. Where did you pull that number from?
If you want other statisticians to take you seriously, you do. I learned it from my professors.
Let's say we have 100 subjects (50 cases, 50 controls), and 100 tests that we're performing. Let's say one of those tests segregates perfectly with our subjects. Bam! Huge result statistically.
Or maybe poor experimental design? Were the cases truly identical to the controls? In a clinical trial, how do you find 50 cases and 50 controls who are exactly alike? How do you ensure that they really underwent identical treatment, as opposed to overly eager doctors and nurses who influenced the outcome? "Double blind" trials, in the real world, are a sham -- doctors often find out who the controls and treatment patients are anyway, based on "safety" excuses.
The probability of that result is astonishingly small (p ~ 10^-30). That means unless we are performing close to 10^30 tests, we can believe this result.
Or did the researchers exclude data they didn't like, to make the p-value small? Are we sure that genetics researchers are not under huge pressure to come up with significant results, given their scarce grants?
First, negative results are in fact published in this field.
Very rare. How often do we see such articles on, say, Science magazine? Scientific American? Slashdot? Or even the Journal of Genetic Epidemiology?
See here for an example about this very study!
How sure are they that the locus in question really came from 18q22? One of the dirty secrets of this field is that people are less sure about which chromosome the DNA came from than they care to admit.
or here for another diabetes replication.
So they passed data through two cohorts, and found several loci, "emphasizing the contribution of multiple variants of modest effect" -- a euphemism for "noise".
All the single-gene disorders with high penetrance have been known since a long time ago. Discovering the genetic origin of diseases such as hypertension or diabetes with (assumed) polygenic origins, which include (assumed) environmental interactions, remain a fantasy that has been "just a few years away" for a long time. In all likelihood, many different mechanisms underlie each disease. While some forms of mental retardation such as Trisomy-21 are easy to test for, there are dozens of other forms of mental retardation with uncertain genetic origins, possibly not genetic at all. Of course, there are other diseases with known genetic mechanisms, such as Huntington's Disease, which is unlikely to occur until old age -- should babies be screened for that? Should kids be told to expect short lives, after which they will become drooling basket cases?
Speaking as a statistical genetics insider, I can tell you that the replication of results in this field is very poor. A team of scientists somewhere will announce they found a gene for XYZ, which is reported in scientific journals and mainstream media -- however, the findings fail to be replicated by other scientists, and the negative results are usually not published. Over the years, hundreds of scientists have claimed to find genes responsible for diabetes, hypertension, autism, etc.
Since there are tens of thousands of genes in the genome, a study with 17,000 subjects makes for less than one subject per gene. (Exactly how many "genes" are in the genome anyway? What exactly defines a "gene"? That is another vast topic.)
Statistically speaking, there should be at least ten subjects per covariate (gene) tested. There is a great deal of hoo-haa over microarrays, but the more you learn about microarrays, the more you will learn just how unreliable they are. The same "disease" can have vastly different pathophysiologies and genetic origins across population groups. Epigenetics, penetrance, expressivity, intron effects -- all multiply the complexity exponentially.
In short, genetics is to biology what nuclear fusion is to physics -- a promising technology that will remain a "few years away" for decades to come.
For decades, medical scientists were sure that obesity raises cancer risk. However, a slew of large recent studies have cast this theory into serious doubt -- obesity has little, if any, statistical association with many cancers. I would take this study with a big grain of salt.