My Thoughts On the Heritability of Intelligence and Eugenics

  • Thread starter Apollo
  • Start date
  • #26
912
0


Originally posted by Nereid
SelfAdjoint: "I think Nachtwolf is quoting from the Flyn book "IQ and the Wealth of Nations". There's a lot of misinformation about this book, such as he did his own tests, used Raven Matrices, etc. Actually it was all done off of published data with little or no attention to different methodologies in different countries.

Neither Apollo, nor Nachtwolf, nor hitssquad have provided any data other than that in Lynn's book, nor offered a counter to SelfAdjoint's characterisation of that book.
The point, from a statistical worldview, is not whether Lynn personally produced the national IQ data, but whether there was enough of that data available to develop quantifiably reliable and statistical trends, and whether major contingencies have been named and statistically quantified. Both the reliability and validity of the data were quantified and the relevance of those quantification further rest upon the contingencies of their own reliabilities and validities. The fact that statistical tools allow us to quantify both reliability and validity of data means that we can develop statistical inferences of the meaning of what Lynn and Vanhanen have brought to us, without having to go frame-by-frame over Zapruder-type films of the original data collection procedures, and without collecting things like standard deviations of the raw scores in each original IQ data collection academic article.

Quoting a previous message in this thread by hitssquad
The reliability of these measures was
quantified by the correlation between two measures taken from a number of countries. In the sample of 81 nations, there are 45 for which there are two or more measures of IQ. There are also 15 countrues for which there are two or more measures and for these we have used the two extreme values. The correlation between the two measures of national IQ is .939. This high correlation establishes that the measure of national IQ has high reliability.
(Lynn, R, and Vanhanen, T. IQ and the Wealth of Nations. p64.)
The validity was quantified.
The validity of national IQ measures was quantified by comparing IQ test results with popular assessments in those nations of bright and dull. It turns out that assessment of bright and dull is universal among all human cultures and that these assessments have strong correlations with IQ test results.
This has been shown to be true in Turkey (Kagitcibasi and Savasir, 1988) for Ugandans, Eskimos, and Native American Indians (Hakstian and Vandenberg, 1979) and for Blacks as well as Whites in South Africa (Kendall, Verster, and von Mollendorf, 1988).
(IQ and the Wealth of Nations. p65.)


Two other national IQ validity assessments consisted of comparisons with reaction time scores and comparisons of national education attainments. For the reaction time comparisions, r scores ranged from .73 to .96, with the higher r scores corresponding with higher statistical significance. For the education attainment comparisions

The correlation between national IQs and mathematics achievement scores is .881 (N-38) and between national IQs and science achievement scores .868 (N=38).... All countries are relatively close to the regression line which indicates a very strong correspondence between national IQs and national differences in school achievements.
(IQ and the Wealth of Nations. p71.)

In a statistical worldview, statistical reliability and statistical validity matter. These things are focused on as important.

Originally posted by Nereid
Apollo, Nachtwolf, hitssquad: if you want PF members and guests to take what you post seriously, please do us the courtesy of answering simple, basic questions about the research on which your assertions appear to rest.
Some of the specifics you have been asking about do not seem to be relevant in a statistical sense. Whether or not Lynn's IQ regression line is flawed, in a statistical sense, is quantifiable by measures of reliability and validity. Lynn and Vanhanen provided a reliability figure of .939; and validity figures of reaction time correlation with IQ r values from .73 to .96, a national mathematics achievement score correlation with IQ r value of .881, and a national science achievement score correlation with IQ r value of .868.

From the perspective of a statistical worldview, the reliability and validity of Lynn's and Vanhanen's IQ/Wealth regression correlation work is contingent upon the reliability and validity of those reliability and validity measures.


-Chris
 
  • #27
selfAdjoint
Staff Emeritus
Gold Member
Dearly Missed
6,786
7
There are also 15 countrues for which there are two or more measures and for these we have used the two extreme values. The correlation between the two measures of national IQ is .939. This high correlation establishes that the measure of national IQ has high reliability.

No, it just shows the same causes produced both values. If those causes were systematically skewed, the same high correlation would result. BTW, which countries were those 15?
 
  • #28
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
hitssquad wrote: Without information on the test protocol, we cannot use the test results to do what?
To eliminate systematic errors (N tests may yield similar results, with high correlations; if all tests suffer from the same systematic error, conclusions drawn from the correlations may be wrong); to address challenges to the validity of the test results; to rule out the null hypothesis.

In short, to conduct a study based on the scientific method.
 
Last edited:
  • #29
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
hitssquad wrote: *SNIP And similarly to what I wrote in another message, if a population tends to be tired -- as with a tire that, perhaps unbeknownst to you, has a nail in it and tends to leak air fast -- and this tiredness affects IQ levels, and it is IQ levels you are looking for since you perhaps want to determine how much variance IQ levels account for in economic variance, then -- as with the air-pressure of the tire -- you would want the level of IQ as it comes from whatever sources of IQ variance the population is afflicted with, not absent them.
Or, in other words, the sub-Saharan African nations' average IQ of 70 may be due to general, population-wide malnutrition; endemic disease; colonial repression (many of the tests Lynn appears to have used were taken during post-WW II independence struggles); or a dozen other factors as yet undetermined ... and NOT RACE [?] [?] [?]

And why is this? Because Lynn et al appear to have not even tried to determine the size (or even the existence) of these potential systematic effects.
 
  • #30
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
hitssquad,

You have made a number of statements concerning the power of statistical analyses, and how conclusions can validly be drawn using such analyses. You use such common statistical terms as 'correlation', 'variance', 'regression', 'significance'. You quote from Lynn, giving values for r, and (sometimes) sample sizes.

Yet on two fundamental aspects - standard deviation and observed distribution about the mean - you have remained silent.

Perhaps the social sciences are different from physics and astronomy, chemistry and engineering? In the latter disciplines, AFAIK, any serious work which involves statistical analysis must include at least consideration of the observed distribution and the standard deviation. You may read that the Hubble constant is 70 km/sec/Mpc, and that a careful analysis of the errors (which includes the observed distribution about the mean and the standard deviation) gives 1-sigma values about this value of 70 of -12 and +15.

Results quoted by you, Apollo, and Nachtwolf have been given as two digits, e.g. 68, 80; or three, e.g. 100, 106. From this one could guess that the errors - however defined - are approx 1% to 5%. However, a proper statistical analysis of the errors may show that the numbers quoted are +/- 12, or 25, or 46 (Nachtwolf, in one of his posts, hints that a difference of 2 is within the precision of the study, but that 6 is significant). From what you (and Apollo, and Nachtwolf) have posted, it is not possible to say whether even the simple statements you make appear reasonable, let alone do our own analyses to confirm or contradict your assertions.

Can we have some standard deviations and observed distributions about the (stated) means please?

Nereid
 
  • #31
912
0
Reliability vs validity

Originally posted by selfAdjoint
hitssquad wrote
There are also 15 countrues for which there are two or more measures and for these we have used the two extreme values. The correlation between the two measures of national IQ is .939. This high correlation establishes that the measure of national IQ has high reliability.
No, it just shows the same causes produced both values.
In a statistical sense, there can be neither a determinable thing as a cause, nor production, of an event.

You seem to be answering here something that was not stated, selfAdjoint. Relative validity was not claimed here; relative reliability was.


If those causes were systematically skewed, the same high correlation would result.
...Which would make the results reliable. This essentially what Lynn wrote. Here it is again: "This high correlation establishes that the measure of national IQ has high reliability." Whatever is causing this skew, as far as our statistical methods have statistical power, appears to be relatively consistent. In a statistical sense, that is all that reliability establishes.

Do you know, in a statistical sense, what reliability is, selfAdjoint? Here's a hint: it is not validity.


BTW, which countries were those 15?
Sorry, my quote from the Lynn book was a typo. After Lynn says there are 45 nations from which there are two or more IQ measures, he says, "There are also 15 countries for which there are more than two measures" (emphasis mine). So, if you want only these 15 nations from which we have more than two measures (as listed in the book IQ and the Wealth of Nations, they are:

Australia
Belgium
Brazil (4 measures)
China
Congo (Zaire)
France
Germany (9 measures)
Hong Kong (5 measures)
India (4 measures)
Japan (10 measures)
Mexico
The Netherlands
South Africa (4 measures)
Switzerland
Taiwan (4 measures)
United States (4 measures)

(If not stated, then the number of measures for a particular nation is three.)


As you can see, I counted 16 nations. I think that perhaps Lynn was not including South Africa since the measures used tended to be racially restricted and the ultimate IQ of South Africa for Lynn's book was an average based on these results and the approximated racial distribution of the society.


-Chris
 
  • #32
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
hitssquad wrote (re conclusions from Lynn's statements re the relative reliability of multiple datasets for the one country): In a statistical sense, there can be neither a determinable thing as a cause, nor production, of an event.
So, in contradiction to what both Apollo and Nachtwolf wrote, you assert that Lynn et al's work merely demonstrates a correlation, but is silent on any causal relationships?

Further, unless and until there is solid work which goes beyond 'a statistical worldview', no hypotheses - including the null hypothesis - can be shown to be inconsistent with the data?

Nachtwolf, for avoidance of doubt, please note these are questions.
 
  • #33
selfAdjoint
Staff Emeritus
Gold Member
Dearly Missed
6,786
7
...Which would make the results reliable. This essentially what Lynn wrote. Here it is again: "This high correlation establishes that the measure of national IQ has high reliability." Whatever is causing this skew, as far as our statistical methods have statistical power, appears to be relatively consistent. In a statistical sense, that is all that reliability establishes.

Do you know, in a statistical sense, what reliability is, selfAdjoint? Here's a hint: it is not validity.


Show me a text where reliability is defined to include effects of systematic bias.

There is (we used to say in the service) precision, and there is accuracy. Precision is when you shoot a two inch circle. Accuracy is when that circle intersects the bull's eye.
 
  • #34
912
0
Coexistence or otherwise of bias and reliability

Originally posted by selfAdjoint
...Which would make the results reliable. This essentially what Lynn wrote. Here it is again: "This high correlation establishes that the measure of national IQ has high reliability." Whatever is causing this skew, as far as our statistical methods have statistical power, appears to be relatively consistent. In a statistical sense, that is all that reliability establishes.

Do you know, in a statistical sense, what reliability is, selfAdjoint? Here's a hint: it is not validity.


Show me a text where reliability is defined to include effects of systematic bias.
First of all, we have as a text your very own quote, reproduced below, which defines reliability as including effects of bias:

There is (we used to say in the service) precision, and there is accuracy. Precision is when you shoot a two inch circle. Accuracy is when that circle intersects the bull's eye.
A non-bull's-eye-intersecting shot group is by statistical definition biased. Precision is reliability is measurement error -- the difference between a tight shot group and a widely-scattered shot group; e.g., how close one might be able to expect a single given shot to conform to the statiostical mean of the shot group -- and accuracy is validity -- e.g., how close that statistical mean gets to the theoretical target.

Precision/reliability, to be an internally-consistent statistical concept, necessarily would have to refer to the relative typical conformity of individual shots to a biased mean of a given shot group of which those individual shots are considered members.

Accuracy/validity, to be an internally-consistent statistical concept, necessarily would have to refer to the relative amount of bias incorporated in the shot group and quantifiable by the relative distance of that shot group's biased mean from any given "nonbiased" bull's-eye.


Arthur Jensen's The g Factor (1998) features the word "reliability" on 83 of its 652 pages. I haven't checked all of those uses of the word within that book, but I would predict that every one of those uses would be consistent with a definition of statistical reliability that assumes that the shot-group mean about which individual shots are judged for relative reliable incorporates an inherent non-zero bias.

Let's look, in The g Factor, at the first instance of use by Jensen of the word reliability:
Spearman noted that the Galtonian measures had poor reliability; that is, the same subject obtained different scores when the test was repeated. Besides measuring what they were intended to measure and appeared to measure, therefore, the Galtonian tests also contained a lot of measurement error. Spearman drew the analogy of firing a gun repeatedly while aiming at a mark on a target. The bullets scatter randomly around the mark, more of them hitting nearer the mark than farther away from it, and the more shots that are fired, the greater is the number of bullets that hit the mark. The scatter of bullets around the mark is analogous to measurement error. It is a part of every kind of measurement, to a greater or lesser degree, depending on the nature of the measuring instrument, the thing being measured, and how hard the experimenter works to reduce measurement error.

Neither Galton nor anyone else working on the measurement of mental abilities had taken into account the reliability of their measurements. Measurement error necessarily diminishes (the technical term is attenuates) the correlation coefficient. The larger the error (that is, the lower the reliability) in either one or both of the correlated variables, the lower will be the possible obtained correlation between them, because the measurement errors are by definition random and therefore uncorrelated. Even two variables that theoretically are perfectly correlated, such as the diameter and the circumference of circles, will not show a perfect correlation (i.e., r = +1.00) unless both variables are measured with perfect accuracy. Yet perfect accuracy of measurement is a pure abstraction never attained by any actual measurement. Actual measurements of any kind always have some "margin of error."

Spearman's formalization of this idea with respect to test scores is the basic postulate of what is now called classic test theory. It states that any and every actual (also termed obtained) score (or measurement), call it X, is composed of two elements--a true score, t, and a random error of measurement, e. (Neither t nor e can be directly observed.) Thus X = t + e. Because e can have either a positive or a negative sign and because it is random, its value tends toward zero as we average more and more of the measurements of X. Theoretically, the average of an infinite number of Xs contains zero error; it will consist purely of t. The way, then, to reduce measurement error is to average a number of repeated measurements of the thing being measured, and to include in the average as many repeated measurements as necessary to achieve the desired degree of accuracy, or, as it is termed, reliability, of the composite measure. The t (which is systematic) is repeatedly averaged in, while the e (which is random) is increasingly averaged out.
In other words, "reliability" is simply an inverse measure of the volume of a stack of bias so complex and individually generally orthogonal to each other that the net sum of those biases is zero with respect to. And, again, the mean of that reliability would consistently, in a statistical sense, have to itself be subject to a non-zero bias. (In other words, the statistical noise of which reliability is categorically and wholly an inverse measure is bias itself, though bias that cancels itself out. Any bias which does not cancel itself out is systematic bias and therefore causes a mean of a shot group to deviate {unlike "noise" bias which merely effects individual shot but not the overall meam of the shot group}.)

So, the "pure noise" bias of which reliability is an inverse measure is simply bias that does not systematically result in deviation from a given golden standard. Out of that "pure noise" bias -- which is to say statistically self-cancelling bias -- we rate with another inverse measure known as validity the relative strength of any non-pure-noise bias.

In short, the mean of any shot group of any amount of reliability necessarily always incorporates a non-zero level of bias which causes a consistent drift from a given golden (ideal; by definition, non-biased) standard.


*edit: format fixed*


-Chris
 
Last edited:
  • #35
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
hitssquad wrote: Sorry, my quote from the Lynn book was a typo. After Lynn says there are 45 nations from which there are two or more IQ measures, he says, "There are also 15 countries for which there are more than two measures" (emphasis mine). So, if you want only these 15 nations from which we have more than two measures (as listed in the book IQ and the Wealth of Nations, they are:
Australia
Belgium
Brazil (4 measures)
China
Congo (Zaire)
France
Germany (9 measures)
Hong Kong (5 measures)
India (4 measures)
Japan (10 measures)
Mexico
The Netherlands
South Africa (4 measures)
Switzerland
Taiwan (4 measures)
United States (4 measures)
If this is so, he does not use all the data in his analysis. Specifically, the appendix contains only the following (I simply clicked on your link):
Australia (2)
Belgium (1)
Brazil (4)
China (1); unless you count Hong Kong and Taiwan as China (which would be sensible, but Lynn doesn't seem to have done this)
Congo (Zaire) (1)
France (1)
Germany (2)
Hong Kong (3)
India (3)
Japan (1)
Mexico (3) YES!
The Netherlands (2)
South Africa (1)
Switzerland (2)
Taiwan (2)
US (1)

It would be interesting to know if he has introduced even more systematic errors than I have already identified by selective inclusion of data! I'll check when I have more time.

[Edit: I just noticed this quote; hitssquad of Lynn (my emphasis): "quantified by the correlation between two measures taken from a number of countries. In the sample of 81 nations, there are 45 for which there are two or more measures of IQ. There are also 15 countrues for which there are two or more measures and for these we have used the two extreme values. The correlation between the two measures of national IQ is .939. This high correlation establishes that the measure of national IQ has high reliability." Strong contradiction between the two sets of Lynn quotes, for six countries (Belgium, China, France, Japan, South Africa, US); slightly less strong for another four (Brazil, Hong Kong, India, Mexico).]
 
Last edited:
  • #36
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
Lynn's work has obvious systematic errors

hitssquad wrote: The point, from a statistical worldview, is not whether Lynn personally produced the national IQ data, but whether there was enough of that data available to develop quantifiably reliable and statistical trends, and whether major contingencies have been named and statistically quantified. Both the reliability and validity of the data were quantified and the relevance of those quantification further rest upon the contingencies of their own reliabilities and validities. The fact that statistical tools allow us to quantify both reliability and validity of data means that we can develop statistical inferences of the meaning of what Lynn and Vanhanen have brought to us, without having to go frame-by-frame over Zapruder-type films of the original data collection procedures, and without collecting things like standard deviations of the raw scores in each original IQ data collection academic article.
For reasons I posted elsewhere, I agree it's not necessary; there is plenty of evidence of failure to account for obvious bias and systematic error in Lynn's work, using nothing more than the data he himself quotes. If you, or anyone else, wants to do a more rigourous study, based on the data Lynn says he used, I would urge you to do "go frame-by-frame over Zapruder-type films of the original data collection procedures", calculate the sample distributions, measures of their scatter, etc.
 
  • #37
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
hitssquad wrote: The validity of national IQ measures was quantified by comparing IQ test results with popular assessments in those nations of bright and dull. It turns out that assessment of bright and dull is universal among all human cultures and that these assessments have strong correlations with IQ test results.
This may well be true; it establishes that within a group the IQ test instruments match - somewhat at least - folk's "assessments of bright and dull".

However, there's a corollary which you might consider interesting: you don't need to use SPM or CPM to determine IQ levels; a simple verbal questionnaire designed to elicit intuitive feelings of 'brightness' and 'dullness' will do just as well. Courtesy of Lynn et al, the reliability of this simple test has already been established.

Perhaps we could enlist PF members in different countries to do this for us? We wouldn't even need to tell them how to select their samples (just record the ages of the subjects); after all, didn't you say that details of the test protocol are irrelevant?

Further, we needn't worry that, in making a score for each subject's 'bright/dull' index, we didn't consider the distribution of the numbers about the mean, nor the SD of that distribution; didn't you already tell us that, in a statistical worldview, it's sufficient that we find a signal?

Finally, the fact that Nachtwolf normalised his test to 100 in the US, that Moni used 200, and Monique 2.3 is also irrelevant; as long as we can show that there's a strong correlation between different PF members' "National bright/dull index" numbers within a country, we're OK (it also doesn't matter that Monique is the only one who does studies in the Netherlands (and she does a dozen), and thehey just one in China).

Wait a minute, I forgot; unless the results show that the sub-Saharan Africans are, as a group, relatively dull, and the east Asians relatively bright, Apollo and Nachtwolf may not believe the answers. Should that concern you? [b(]
 
  • #38
912
0
The prerogatives of the researcher and why they are his prerogatives

Originally posted by Nereid
However, there's a corollary which you might consider interesting: you don't need to use SPM or CPM to determine IQ levels; a simple verbal questionnaire designed to elicit intuitive feelings of 'brightness' and 'dullness' will do just as well.
If it would prove to be predictive, then it would prove to be predictive. He could have also chosen reaction-time tests and/or achievement tests and/or sought the aid of psychics.

However, experience with human cognitive faculty measures -- both IQ and non-IQ -- in the United States shows that they are only as generally predictive in terms of sociologically important outcomes as they are loaded on the g factor of mental ability. It is the researcher's prerogative to use whatever methods he chooses. If he chooses wrong, then he ends up with a source of variance that proves to be not very predictive (and some academic publishings documenting his failure to competently theorize). Reaction-time tests and achievement tests are not as highly g-loaded as IQ tests (and I don't know off the top of my head what the g-loadings of psychic estimates of IQ typically are). One might imagine that Lynn chose IQ tests because they are the most-g-loaded measures of human mental faculties and therefore can be expected to be the most predictive and the most helpful in successfully reaching the goal of establishing a new explanatory theory.



-Chris
 
  • #39
Njorl
Science Advisor
258
10
I remember a study done to determine brain volumes of whites and blacks. It was a long time ago. The researcher filled skulls with beans, then counted the beans. He determined that whites' brains were much larger than those of blacks. A few years later, another researcher obtained the same skulls and used water rather than beans. He found no statistically significant difference in skull volume between the races.

Quite a bit of research has been done by people with agendas - justifying slavery, justifying colonialism, justifying racism. Studies that rely upon extant data are likely, even though done with innocent motivation, to be contaminated by this.

I doubt that much original basic research of this nature gets done. It is most likely, that different studies are done of the same old data, of dubious origin. It is unlikely that a researcher is going to be discredited by someone repeating his experiments.

I am also very dubious of "culurally neutral" IQ tests. When an aborigine invents one, I'll be a little more inclined to believe they exist. Even more difficult than devising such a test would be administering it in a culturally neutral way. Also, have any studies been done concerning racial attitudes and the administration of IQ tests? What I'm asking is, have any studies been done on the test givers, such that they are secretly the subject of the test to see if their prejudices affect the results of the IQ test?

Njorl
 
  • #40
selfAdjoint
Staff Emeritus
Gold Member
Dearly Missed
6,786
7
Njorl, this is just systematic denial. No research is being done (after all those extracts hitsquad posted), it's all just rehashing old data (proves you haven't looked), everybody has a (bad) agenda (smear the messenger), no test can be culture-free (where is the culture value in Raven matrices)? and so on. You don't want to look at the good science and good data upon which these facts are based so you just color them ugly in your mind.
 
  • #41
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2


Originally posted by hitssquad
If it would prove to be predictive, then it would prove to be predictive. He could have also chosen reaction-time tests and/or achievement tests and/or sought the aid of psychics.

However, experience with human cognitive faculty measures -- both IQ and non-IQ -- in the United States shows that they are only as generally predictive in terms of sociologically important outcomes as they are loaded on the g factor of mental ability. It is the researcher's prerogative to use whatever methods he chooses. If he chooses wrong, then he ends up with a source of variance that proves to be not very predictive (and some academic publishings documenting his failure to competently theorize). Reaction-time tests and achievement tests are not as highly g-loaded as IQ tests (and I don't know off the top of my head what the g-loadings of psychic estimates of IQ typically are). One might imagine that Lynn chose IQ tests because they are the most-g-loaded measures of human mental faculties and therefore can be expected to be the most predictive and the most helpful in successfully reaching the goal of establishing a new explanatory theory.
Seems my idea of a corollary was posted in the wrong thread; it's relevant to the one discussing Lynn's work, not here (or at least, not obviously relevant).
 
  • #42
912
0
The MRI-confirmed relationship between brain size and IQ

Originally posted by Njorl
I remember a study done to determine brain volumes of whites and blacks. It was a long time ago. The researcher filled skulls with beans, then counted the beans.
"...And perhaps he packed the beans more tightly in the black skulls than in the white skulls," goes the anecdote (roughly) from Stephen J. Gould's Mismeasure of Man. 19th-century brain-volume studies are neither necessary to confirm a relationship between brain size and IQ within races nor between races. MRI studies confirm brain size differences correlate with IQ both within and between races.

If you subscribe to Questia (the online library) and search for the term <MRI> in Arthur Jensen's The g Factor (which Questia has in its collection), you will find Arthur Jensen's summaries of the data confirming brain volume correlations with IQ:
http://www.questia.com/PM.qst?a=o&d=24373874



Quite a bit of research has been done by people with agendas
The scientific process filters out agenda. Either the results of research prove to be reliably predictive, or they do not. MRI brain volume measurements have been established as reliably predictive of IQ.

---
direct measurements of in vivo brain size obtained by magnetic resonance imaging (MRI) show an average correlation with IQ of about +.40 in several studies based on white samples. Given the reasonable assumption that this correlation is the same for blacks, statistical regression would predict that an IQ difference equivalent to 1¦Ò would be reduced by 0.4¦Ò, leaving a difference of only 0.6¦Ò, for black and white groups matched on brain size. This is a sizable effect. As the best estimate of the W-B mean IQ difference in the population is equivalent to 1.1¦Ò or 16 IQ points, then 0.40 ¡Á 16 ¡Ö 6 IQ points of the blackwhite IQ difference would be accounted for by differences in brain size. (Slightly more than 0.4¦Ò would predictably be accounted for if a hypothetically pure measure of g could be used.) Only MRI studies of brain size in representative samples of each population will allow us to improve this estimate.
---
p442
http://www.questia.com/PM.qst?a=o&d=24373874

More-recent MRI studies show even higher brain volume correlations with IQ in certain areas of the the brain, such as the frontal lobes.
http://groups.yahoo.com/group/e-l/files/Genetic influences on brain structure.pdf



I am also very dubious of "culurally neutral" IQ tests.
There are no "culturally neutral" IQ tests. There are culture-reduced tests, such as the Raven's Matrices. There are culture-loaded tests that cross cultures with high reliability and validity when they are rewritten to conform to local culture. Reliability and validity of cross-cultural IQ tests is established via statistical methods.



When an aborigine invents one, I'll be a little more inclined to believe they exist.
Any aborigine-invented IQ tests can have their reliability and validity quantitatively established just as any other IQ test can.



have any studies been done on the test givers, such that they are secretly the subject of the test to see if their prejudices affect the results of the IQ test?
Yes. Arthur Jensen visited this topic in his 1980 book Bias in Mental Testing.
https://www.amazon.com/exec/obidos/tg/detail/-/0029164303/&tag=pfamazon01-20




-Chris
 
  • #43
Njorl
Science Advisor
258
10
[q]The scientific process filters out agenda. Either the results of research prove to be reliably predictive, or they do not. [/q]

This is certainly not always true. Vast amounts of scientific rubbish get published and unrefuted frequently. A researcher with an agenda can certainly publish studies with slightly fraudulent data without being detected. There is no monetary incentive to seriously attempt to disprove this work. While there is a lot of money in selling books saying that blacks are inferior, there is no money in proving that they are not. This type of research is expensive and time consuming. Why would someone do the work necessary to refute these ideas.

To claim that these studies stand up to scrutiny, there would need to be a significant community interested in rigorously answering these questions. I do not believe such a community exists. They can analyse methodology, and debate conclusions, but they can not recreate the experiment. I do, however, believe there is a community interested in propagandising racist attitudes. While I'm not prepared to call someone of Jensen's stature a racist, I am also not willing to accept his results until they are duplicated from the ground up by someone who will not profit by the results.

My best friend taught calculus for four years in a small town in Swaziland. His students were 15-17 years old. They were not-preselected for their ability, if anything, they were preselected for their poverty (Swaziland at the time, though, was fairly prosperous compared to their neighbors). The majority of the students had no trouble with the material. This seems inconsistant with an average national IQ of <70. Calculus is not simple. If you have an IQ <80 you will not get it. Personal anecdotes are no substitute for systematic study and statstical analysis - I'm not one of those "Lies, damn lies and statistics" people - but anecdotes are a reason for raising questions.

Njorl
 
  • #44
Njorl
Science Advisor
258
10
Originally posted by Apollo

I am planning to research the databank at http://www.theoccidentalquarterly.com/ which appears to contain a large volume of scientific research on the topics of intelligence, personality, human interaction, race/ethnicity, and human evolution. If I come across any interesting data, I will post them here.
You're not going to have much credibility if you do. This site is essentially the KKK with a college education.

Njorl
 
  • #45
912
0
The minimum mental age required to learn calculus

Originally posted by Njorl
My best friend taught calculus for four years in a small town in Swaziland. His students were 15-17 years old. They were not-preselected for their ability, if anything, they were preselected for their poverty. The majority of the students had no trouble with the material. This seems inconsistant with an average national IQ of <70. Calculus is not simple. If you have an IQ <80 you will not get it.
Lynn assumes a Swaziland national IQ of mean 72. If we assume a standard deviation of 10.8 IQ points (72% of 100), and a normal distribution, 2.3% of the Swaziland population has an IQ above 93.6 IQ points. The age-zero-to-14 population is currently 480,000. A three-year period of that would be one-fifth, which is about 100,000. 2.3% of that 100,000 is 2,300.

At any given time, in Swaziland, there are about 2,300 15-to-17-year-olds ready for calculus, if an IQ of 93.6 relative to a British mean of 100 is sufficient for learning calculus.



-Chris
 
  • #46
selfAdjoint
Staff Emeritus
Gold Member
Dearly Missed
6,786
7
That doesn't fix the problem with the class. Of course we don't know the standard deviation of IQ in Swaziland, but just doing binomial statistics suggests the probability of finding a random sample of say 20 kids with IQ's all 23 or more points above the population mean is astronomical. Kind of falsifies the stated mean, don't you think?

Let's say the standard deviation is 15, like the white polpulation of the US. then z = 23/15 = 1.53, so p= .063. That's for one kid.

.063^20 = 9.7 X 10^-25, for 20 kids.

If you make the s.d. smaller, it just makes the sample more unlikely.
 
  • #47
912
0
Quantities of IQ studies per nation in Lynn's regression analyses

Originally posted by Nereid
hitssquad wrote: Sorry, my quote from the Lynn book was a typo. After Lynn says there are 45 nations from which there are two or more IQ measures, he says, "There are also 15 countries for which there are more than two measures" (emphasis mine). So, if you want only these 15 nations from which we have more than two measures (as listed in the book IQ and the Wealth of Nations, they are:
Australia
Belgium
Brazil (4 measures)
China
Congo (Zaire)
France
Germany (9 measures)
Hong Kong (5 measures)
India (4 measures)
Japan (10 measures)
Mexico
The Netherlands
South Africa (4 measures)
Switzerland
Taiwan (4 measures)
United States (4 measures)
If this is so, he does not use all the data in his analysis. Specifically, the appendix contains only the following (I simply clicked on your link):
Australia (2)
Belgium (1)
Brazil (4)
China (1); unless you count Hong Kong and Taiwan as China (which would be sensible, but Lynn doesn't seem to have done this)
Congo (Zaire) (1)
France (1)
Germany (2)
Hong Kong (3)
India (3)
Japan (1)
Mexico (3) YES!
The Netherlands (2)
South Africa (1)
Switzerland (2)
Taiwan (2)
US (1)

Your data is from the article on Lynn's website.
http://www.rlynn.co.uk/pages/articles.htm [Broken]

The book IQ and the Wealth of Nations is different from that article. I assembled my data from that book's Appendix 1: the Calculation of National Intelligence Levels. pp197-225.


Strong contradiction between the two sets of Lynn quotes, for six countries (Belgium, China, France, Japan, South Africa, US); slightly less strong for another four (Brazil, Hong Kong, India, Mexico).
Your conclusion of contradictions doesn't seem to be supported by my quotes of Lynn's methodology from Appendix 1 of IQ and the Wealth of Nations.


*edit: added title*


-Chris
 
Last edited by a moderator:
  • #48
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
moving target???

Hey hitssquad,

you were the one who posted links to Lynn's methodology, data, conclusions etc here (not me).

I merely used your stated data sources, methods, etc to show that Lynn's study contains serious contradictions, flaws, etc.

Now you're telling us that the sources you quoted aren't correct, and that the real study is elsewhere, used different methods, data, sources, ... and even came to quite different conclusions???

Put yourself in my position - would you get angry? would you feel that you've treated this forum (and its members) with disdain?

If you want to have a proper discussion, based on the accepted methodology of science, by all means please go ahead and start a thread.

However, don't be surprised if you are challenged as to whether you are being honest and straight.
 
  • #49
912
0


Originally posted by Nereid
the real study is elsewhere, used different methods, data, sources,
The online article...
http://www.rlynn.co.uk/pages/articles.htm

...appears to use similar overall methods. The individual details of IQ adjustment in the case of each individual nation are not given in the online article, but they are given in the book (IQ and the Wealth of Nations). The book also goes further in the respect of including several runs through several variations of different timeframes all the way back to 1820, to show how IQ/GDP correlations compare throughout the span of the modern era.

As for the sources, more are listed for the book than for the online article. However, at least some of the online article's spurces may be related to the book's sources. As you have seen, the Lynn study for Ethiopia was actually an article he wrote correcting conversion errors in the study listed as the reference in the book. Both are academic "sources" of the same raw IQ-sampling data.



... and even came to quite different conclusions?
The online article presents the same major conclusion as that of the book, which is that -- in our modern era, at least -- national IQ predicts national income better than any other single measurable factor.


The online article puts this as:

---
The results are interpreted in terms of a causal model in which population IQs are the major determinant of the wealth and poverty of nations in the contemporary world.
---
http://www.rlynn.co.uk/pages/article_intelligence/1.htm


---
our general conclusion is that national differences in the wealth and poverty of nations in the contemporary world can be explained first in terms of the intelligence levels of the populations;
---
http://www.rlynn.co.uk/pages/article_intelligence/6.htm


The book repeats this conclusion in several forms and in several places. On p183 at the beginning of Chapter 10 (The Future of the Wealth of Nation), it says:

---
In general, although not without exceptions, nations with more intelligent populations have been able to achieve a higher level of per capita incomes than less intelligent nations. This is a major reason why economic inequalities among nations are so great.
---
https://www.amazon.com/exec/obidos/tg/detail/-/027597510X&tag=pfamazon01-20



-Chris
 
Last edited by a moderator:
  • #50
Nereid
Staff Emeritus
Science Advisor
Gold Member
3,367
2
... and what was the response of the non-Jensen folks to Lynn's book? Could you point us to some on-line critiques?
 

Related Threads on My Thoughts On the Heritability of Intelligence and Eugenics

Replies
9
Views
3K
  • Last Post
Replies
7
Views
3K
  • Last Post
Replies
16
Views
2K
  • Last Post
Replies
12
Views
2K
Replies
1
Views
3K
  • Last Post
Replies
9
Views
2K
  • Last Post
10
Replies
238
Views
18K
  • Last Post
Replies
2
Views
2K
Replies
4
Views
2K
Replies
2
Views
2K
Top