Jump to content
Objectivism Online Forum

Knowledge gained from statistics.

Rate this topic


Recommended Posts

Damn, I lost £200
Eh, sounds like the pound really sux.
As far as I'm seeing it terrorists are mostly Muslim who are mostly not white.
But, Shirley, you must see how that is a separate thread. You have implied a correlation between terrorism and Arabicity. I have therefore (irrationally???) assumed that you meant what you said. The sad thing is that you said that amongst other things that ALL of the London bombers were Arabs when in fact NONE of them were, and equally you said (or, in a sneaky, smarmy way implied) that most terrorists are Arabs. By "Arabs", I mean, and I assume you mean, as a speaker of the King's English (currently spoken by a queen) actual Arabs, and not "Blacks, wogs, non-whites, ethnics, Spaniards, whatever". I am not questioning the need to do what is necessary to survive, I'm questioning the validity of making things up because you think it is necessary to survive. If you can prove, please read OPAR to get a grip on that concept, prove, that most terrorists are Arabs, then do so, otherwise STFU about this made-up correlation of Arabicity and terrorism.

People die when statistics are falsified. Do not falsify statistics. That is exactly the tactic of the eco-terrorists: fabricating a representation of reality is utterly abhorrant to Objectivist. The last thing an Objectivist, or Objectivist-wannabe, should consider doing is faking reality because they have an emotional attachment to a conclusion like "All darkies should be strip-searched at the airport because there's a better than zero chance that they are terrorists".

Please, for the love of god, tell me that you actually, finally, get it.

Link to comment
Share on other sites

So a given airport can weight *these* probabilities to get the best course of action.

Just blindly applying Bayes' rule is not enough to effectively design screening protocols for airport security.

First of all, assuming that we properly computed the probability that a randomly sampled terrorist exhibits some phenotype, such as possessing Arabic features is not automatically the same as the probability that a randomly sampled terrorist who is going to board a plane exhibits those same features. The probabilities may turn out to be the same, but this would need to be established.

This distinction is especially important to make because terrorists know that the western governments can undergo naive profiling techniques. Therefore, they would be more likely to recruit atypical extremists such as Richard Reid or Jose Padilla. Given that lives are at risk, it is crucial that any screening process incorporate a game theoretic element.

Edited by DarkWaters
Link to comment
Share on other sites

I've lost count of how many times I've been to the airport.

On one trip I saw security stop a very old lady and let a middle aged man wearing a turban (spelling?) pass, he looked middle eastern. I thought that was ridiculous. I'd like to see statistics on old ladies being terrorists. If any could be found. Is it not reasonable to assume that most terrorist are not old ladies?

At the same time I've seen another man, who looked to be from middle eastern decent, watched very carefully. He was given all sorts of harsh looks. I heard snide remarks from a typical Texan redneck. Yes I just said redneck, sorry... but it's the best way to describe the person. What's worse is that I thought the same thing that the redneck said. I thought that if he did anything I would do my best to physically stop him. I later heard the middle eastern looking man talk and he had an American accent. That made me feel like shit. I felt sorry for man. He didn't deserve to be treated that way. He must have felt very uncomfortable.

This whole subject makes me think of how Japanese Americans were treated during WWII. That was a shame.

Civil Liberties versus Security. Yuck. Touchy subject.

To me it makes sense to focus security screening, but not just on anyone middle eastern or Arabic. I would not want to be the person to create such a law. I would gladly help make a law to focus security on males, particularly young and middle age males, but not completely exclude anyone else.

Link to comment
Share on other sites

By the way I am pretty sure I am not racist. I've dated a stunning and smart black girl before so I do judge an individual on their merits.

This is not meant to be a judgement on you personally as I lack the basis to make such an evaluation. I am instead attacking your reasoning.

Dating a black woman is not sufficient to demonstrate that one is not racist. Misogynists can still have wives and girlfriends. One could even be dating an individual of a different race out of some fetish but not out of respect. Even the former segregationist Strom Thurmond had fathered an illegitimate black daughter.

Truly evaluating all individuals on their merits (as you claim to do) and truly treating all individuals with the respect that they deserve are much more convincing than just claiming to have a friend who is a minority.

Link to comment
Share on other sites

Look, just give me a real statistic that isn't just pure racist rationalism invented because of some bizarre belief that Arabs are somehow inferior, or something like that, and you at least have a statistic. This "there can be a statistic" baloney does not mean that there actually is a statistic. My god, that is the fundamental credo of "primacy of consciousness", to say "if X 'can' exist, it therefore does exist". For my money, this appalling lack of interest in the actual facts of existence is exactly what Objectivism is utterly, totally, and dimetrically opposed to.

I know that you are not a total mathematical ignoramus, so I figure it probably is not necessary to explain elementary statistical methods to you, but are you really such a Platonist that you can't even see how a central assumption in all statistical reasoning is that you actually have valid observations?

I never claimed there was a statistic correlating the Arab race and terrorism. As I explicated before, the question does not hinge on a particular statistic but abstracts about statistics as a whole. I have given you one such statistic involving my house. That it has no globally significant application is not important--because, again, it is an example of addressing statistics qua statistics, not qua politically significant statistics. And nobody said that considering statistics as a whole implies the claim that there is some particular kind of statistic. Where you ever found that in anything that has been written on this topic or in any other is entirely beyond me.

But to provide one statistic that correlates irrelevant facts, but which might be of interest for practical action, would be the correlation between race and crime. Being black doesn't make you commit crime (to say that would be racist), but there is a satistical, numeric correlation between the black race and crime. (http://transcripts.cnn.com/TRANSCRIPTS/0702/07/pzn.01.html)

However, even this should not be necessary. The lack of interest in actual facts is due to intellectual division of labor. There are a great many facts I am not interested in. The various types of soil that exist, and their chemical composition. Modern designs incorporated into upholstery. Methods of hunting used by South-east Asians. None of these interest me much, because I leave it to geologists (or whoever studies soil), interior designers, and sociologists. I leave actual, concrete statistics to politicians, police forces, statisticians, sociologists, psychologists, and whoever else might be interested. The question being proposed here, however, is not about any particular statistic with any particular relevance to any particular study (terrorism, it seems to me, was a hypothetical). Just as we might study numbers without asking which numbers, or numbers of what, we may study statistics abstractly. We leave the "which" and "of what" out of the question--a blank check, to be filled in later when a particular application is needed. That does not mean that the "which" and "of what" cease to exist or apply. In number we may say that x plus x equals 2x. Which x? Well, in a particular instance we may fill in two. Two plus two is two times two. Two of what? Well, in a particular instance we may fill in cows. But for now, we leave those questions alone because they don't concern the general treatment. So here, talking about statistics, we don't need to know which statistics or of what. We treat it generally and leave the "which" and "of what" to be filled in upon demand in any particular instance. I should think, Objectivism placing as much stress as it does on concepts, this should be a very exciting study for some Objectivists.

Funny you should call me a Platonist--my logic professor has just butted heads with me rather forcefully because I do not accept his Platonism, but take a more verificationist and natural view of math and logic.

Link to comment
Share on other sites

In the book Racial Profiling by Deborah Kops, pp. 78-79, she cites a statistic reported by the FBI that, of the terrorist events reported by some of the more populous countries in the world (that is to say, they did not spend their time investigating Andorra's terrorist statistics), the combined groups of Arabs, Persians, and Afghanistanis were responsible for less than three-quarters of the events and more than half the deaths and injuries in the last twenty years.

So now, based on that information, is racial profiling justified?

With that said, the statistics I just wrote are a complete fabrication (don't believe everything you see on the internet). The point is, the conversation asks, "What if we had such a statistic?

I can see that you thought that any reasonable person would read those facts and would find the answer to your question to be an obvious "yes".

Honestly, before my eye went to the next paragraph, where you admit it's a fabrication, I was already thinking: "How can one jump to that conclusion?" It is a huge leap. When I read your example, my first question was: what are they counting as terrorism? Are they counting all the bombings in the middle east? Because if they are, then that's not too relevant outside the middle east.

There may even be places where Arabs are less likely to be profiled. If an Arab flies into Belfast airport, he might find that the cops are less likely to bother about him than they do about a few people who fit their profile of an IRA member. So, what happened to the Arab-causal factor?

Whenever one reads stats like that, one must first try to understand the full context of the statistic.

Link to comment
Share on other sites

Actually, I'm ambivalent about racial profiling. In principle, I don't oppose it, but I'd like to see evidence that it works better than all alternatives. I wrote that as an illustration. I wanted to point out that, for most any reader, before reading that the facts were fabricated they would start seriously entertaining the idea that the facts were true and start considering what the implication of them would be--just as you had.

As for the causal factor, I was illustrating that causality is not the essential concern of statistics. Statistics concerns itself with analysis and comparison of sets of numerical facts.

Link to comment
Share on other sites

I wanted to point out that, for most any reader, before reading that the facts were fabricated they would start seriously entertaining the idea that the facts were true and start considering what the implication of them would be--just as you had.
Unless a reader has cause to doubt the source, he'd assume that the source was reasonably honest. I don't think a reader would assume the facts were true as stated, but would assume that the facts were true in some sense. So, the serious entertainment could simply be of the form "I wonder what skewed this observation" rather than "this guy is lying". A key element is the extent to which the new information integrates with what one already assumes. For instance, if you had written about "Chinese" instead of "Arabs" in your example, more alarm bells would go off. Anyway, I'm not sure where all this gets us.

I have a suggestion to make about the thread, and it is that the best examples to consider as a starting point would be one that are not complex, and ones that do not involve common stereotypes, Those are bound to take the topic elsewhere. I suggest a neutral example along the lines of: Harold has two orchards that he thought are pretty similar, but the fruit from one are usually bigger and juicier than the fruit from the other...now Harold wants to know if it is valid to do XYZ, or something of that sort.

Link to comment
Share on other sites

Well, considering the nature of the internet and especially these posts, I don't think there's any reason to assume factual corroboration with claims made beyond the ability of a poster to cite respected websites like those of news organizations, encyclopedias, etc. I could proliferate "false facts" all day, just doing a quick search in amazon.com for a book title that seems to suit the purpose and listing some random page number (maybe with a chapter heading for kicks, which you can look up in amazon.com's preview images). In fact, one lady took advantage of the gullibility of internet readers and circulated a list of "facts" that she had completely concocted. As a result, it's now a semi-common myth that the average person eats an average of four spiders in their lifetimes.

This not withstanding, my point was entirely different. I was simply making the reader start to hypothesize, as was requested by the original post. "Supposing that such-and-such statistic were true...", "Were such-and-such true..." and the like. By making the reader at least initially consider the statistic possible and start wondering what would follow if it were true, I cleverly forced the goal I had set for myself. :)

Link to comment
Share on other sites

As for the causal factor, I was illustrating that causality is not the essential concern of statistics. Statistics concerns itself with analysis and comparison of sets of numerical facts.

No, but statistics are supposed to help researchers discover causalities and invalidate other claims of causality. Otherwise, what is the point of all of the analyses and comparisons? But you probably agree to this anyway.

Link to comment
Share on other sites

No, but statistics are supposed to help researchers discover causalities and invalidate other claims of causality. Otherwise, what is the point of all of the analyses and comparisons? But you probably agree to this anyway.

There have been some pretty good posts on this topic. I mentioned which ones I thought were especially good in an earlier post on this thread. I will summarize why statistics are a valid cognitive tool, and several ways they are mis-used. The key principle is that statistics establish correlations; they do not establish causation without further (non-statistical) research. Why are statistical correlations so valuable?

First, they form a basis for action, when the need to act is immediate and there is no time to establish causal relationships. Examples of that have been given. Frequently, it is simply not worth the effort to establish causation. In those cases, basing actions on statistical correlations means getting things done that would not otherwise get done.

Second, they can point up areas for further research. Analyzing a set of data using a technique called "regression analysis" can show up a heretofore unknown correlation buried among a seemingly inchoate mass of data. That discovered correlation can then be analyzed for a causal relationship. The result is new knowledge. Statistics give you a tool to tell you when the correlation is "statistically significant", i.e., it is unlikely to be just random chance. A higher degree of correlation indicates a stronger relationship, one that is more worthy of study. Statistics can give you a measure of just how strong the correlation is.

There is no problem with the science of statistics. There are only problems when statistics are mis-used, which happens frequently. Here are some fallacies of statistics that you can see almost every day in the newspapers or on the evening news:

(1) Assuming a causal relationship exists when there is a correlation. Correlated phenomena may be causally related, they may be caused by a third phenomenon, or there may be no causal relationship. Furthermore, the correlation itself could simply be coincidence. Correlation does not necessarily imply a causal relationship, although a stronger correlation indicates that a causal relationship is more likely. To put it simply, correlation alone cannot establish a causal relationship.

For example, not too long ago in the news there was a story saying that teenagers who used anti-depressants were more likely to commit suicide. The FDA ordered drug manufacturers to put warning labels on the anti-depressants, acting under the premise that the anti-depressants caused the teens to commit suicide. Can such a conclusion about causation be derived from a mere statistical correlation? Of course not. In fact, the more likely reason for the correlation is simply that teens with a greater propensity to commit suicide used anti-depressants.

Which conclusion is true cannot be answered from statistics. Further medical research is required. It appears that the FDA acted in hasty reliance upon an improper conclusion drawn from a statistical correlation. As Bart Simpson would say, "Duh!".

(2) Mis-stating the direction of causation. Assuming Phenomenon A caused Phenomenon B, when the reverse is true. For example, a medical researcher may observe that people with a certain cancer are missing a key enzyme in their blood. He assumes that the lack of that enzyme causes the cancer, and proposes that people ingest that enzyme in order to prevent the cancer. However, further medical research showed that causation was reversed. The cancer caused the blood to lose the key enzyme, not the other way around. The research reached a faulty conclusion about the direction of causation by improperly relying only on statistics for his conclusion.

(3) Assuming that A and B are causally related when, in fact, both A and B are caused by a third factor, C. For example, statistics may show a correlation between poverty and poor education. An advocate of public education may propose that free education will bring people out of poverty. He spreads public education far and wide, but the the poor remain poor. He may have missed the fact that both poverty and poor education are together caused by a third factor, mistaken moral values. If a group of people think that it is not "cool" to be "book-learned", that group is likely to end up both poor and poorly educated.

Statistics alone cannot identify where the causal relationships lie. More research is needed. The public education advocate mis-used statistics to reach a faulty conclusion about causation.

(4) Problems with small sample sizes. It is an error to assume that an observed correlation among a small set of data will prove true among the larger population. Conclusions about a larger population are more likely to be true when the sample size is larger. Small sample sizes yield unreliable conclusions about correlation. For example, a survey of 10 Americans about who will win the next Presidential election is worthless as an indicator of current sentiment in the population. However, a survey of 3,000 Americans gives a much more reliable indicator of the current sentiment in the population. Watch survey data carefully to see how many people were surveyed. Often, they are based on sample sizes that are too small.

This is not an exhaustive list of statistical fallacies. Statistics is a very useful tool in many fields, from medicine to stock market investing to engineering. As ubiquitous as it is, it is well worth studying, especially so that you can spot the fallacies, which occur so frequently, and not make them yourselves.

Edited by Galileo Blogs
Link to comment
Share on other sites

Added to all those methodological errors, one also has researchers with agendas and reporters who want to sensationalize things.
My opinion is that this is actually the most significant problem with using statistics to arrive at conclusions. Passive minds simply absorb statistically encapsulated summaries without knowing or even caring how the numbers were derived. Once it's distilled down to a number, we needn't inquire as to whether that number has a relation to reality. It is all too easy to fake a statistic created to support an agenda, and it happens all too often.
Link to comment
Share on other sites

I agree with DavidOdden's and SoftwareNerd's comments. A saying captures those thoughts, and much more:

"Torture the data and Nature will repent."

I first heard it as inside joke from an economics professor of mine, a professor who spent all day every day doing countless regressions to "prove" economic trivialities.

Statistics are a convenient and easy weapon to deceive. That is why there are so many examples of statistics being mis-used that way.

Edited by Galileo Blogs
Link to comment
Share on other sites

There have been some pretty good posts on this topic. I mentioned which ones I thought were especially good in an earlier post on this thread. I will summarize why statistics are a valid cognitive tool, and several ways they are mis-used. The key principle is that statistics establish correlations; they do not establish causation without further (non-statistical) research. Why are statistical correlations so valuable?

::: SNIP :::

This is not an exhaustive list of statistical fallacies. Statistics is a very useful tool in many fields, from medicine to stock market investing to engineering. As ubiquitous as it is, it is well worth studying, especially so that you can spot the fallacies, which occur so frequently, and not make them yourselves.

Fantastic post. I agree with everything you stated and I never thought otherwise.

Some other common statistical fallacies that I can immediately think of include:

Obtaining a biased sample.

Failing to account for collinearity.

Mistaking something that is statistically insignificant for being significant.

Failing to properly remove outliers.

Unsystematically removing unfavorable outliers to make one's results confirm one's hypothesis.

Of course, this list added to yours is also not an exhaustive list as there are probably numerous other assumptions that are required for a statistical test to be valid, such as normality of the distributions within groups, homoscedasticity (homogeneity of the variances) and independence of the observations, that are often overlooked.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...