Falsely reported negatives

(Edit: Corrections were made to this.)

In his post, Science for fun and profit, Steve Sailer notes:

“One thing that’s interesting is how seldom these kind of data-mined false positives are published regarding The Gap, despite the huge incentives for somebody to come up with something reassuring about The Gap.”

What you get at times instead of false positives are negatives that are reported as if they were positives. For example, below are the results from the two published, and not infrequently cited, transracial adoption studies in the US concerning Blacks and Whites, Moore (1986) and the MTRAS, with statistical differences noted. (I calculated d’s using the means and standard deviations presented in the original papers.)

Column F gives the Cohen’s d found between the adopted groups (e.g., Blacks and mixed race kids). Column G gives the significance levels based on Welche’s unpaired t, 2-tail, 0.05. Column H gives the Cohen’s d predicted by a genetic hypothesis with genetic d =1 and a heritability of 0.4 in childhood and 0.6 in adolescence (i.e., 1 x SQRT(.4) and 1 x SQRT(.6)). Column I gives the Cohen’s d predicted by a genetic hypothesis for WGH = 0.5. Column J gives the Cohen’s d between the found results and the genetic prediction in H (i.e., H-F). Column K gives the Cohen’s d between the found results and the genetic prediction in I (i.e., I-F). Column L gives the significance of J (again using Welche’s unpaired t). Column M gives the significance of K (again using Welche’s unpaired t).

…. Sure the Moore (1986) results which are lauded as proof of racial environmental determinism weren’t significantly different from zero, but they also weren’t significantly different from a genetic hypothesis’s predictions (genetic d =1). (And using one tail tests while increased the WGH to 0.5 didn’t remedy the lack of significant effects.)

But no one mentions the latter point. Richard Nibett, for example, in his 2007 NYtimes op-ed, misleadingly informs us:

transracial adoption

Excel file.

“A superior adoption study — and one not discussed by the hereditarians — was carried out at Arizona State University by the psychologist Elsie Moore, who looked at black and mixed-race children adopted by middle-class families, either black or white, and found no difference in I.Q. between the black and mixed-race children. (All Brains Are the Same Color)”

Well, ok.

For comparison the MTRAS — the inferior adoption study which had the deficiencies of being longitudinal, of including multiple indexes of cognitive ability, of including mixed and full race individuals along with adopted and non adopted whites, and of having authors which estimated the heritabilities of IQ in the sample — gives us six adopted estimates: Mixed-Black time 1, time 2 -White-Black time 1, time 2 – White-mixed time 1, time 2. Of then, 5 are significantly different from zero and only 1 is significantly different from the predictions of a genetic hypothesis. (Using one tail tests and setting the WGH to 0.5 for both time 1 and time 2 we get 5 values significantly different from zero and 2 values which are significantly greater than predicted by a genetic hypothesis).

(It follows from the above that, as I noted elsewhere, results from adoption studies — all two of them –provide more evidence for a genetic hypothesis than for a shared environmental hypothesis — no matter how you slice and dice the results. It’s interesting (to me) that no one has bothered to make this point, statistically.)

Now, this is merely one example of misleadingly reported negatives. Others, just from Nisbett’s All Brains Are the Same Color piece, include Nisbett’s discussion of the famed Scarr et al. (1977) study and his discussion of the Witty and Jenkins (1936) study .

So, ya, when it comes to the GAP you won’t get a lot of false positives. But not infrequently you will get falsely reported negatives.

References

Moore, 1986. Family socialization and the IQ test performance of traditionally and transracially adopted black children. Table 2.

Scarr et al., 1992. IQ Correlations in Transracial Adoptive Families

Weinberg et al., 1992. The Minnesota transracial adoption study: A follow-up of IQ test performance at adolescence. Tabel 2.

Advertisements
This entry was posted in Uncategorized. Bookmark the permalink.

10 Responses to Falsely reported negatives

  1. Kiwiguy says:

    That op-ed reminds me of a response I received from someone who attends the International Society for Intelligence Research conferences about Nisbett.

    I would like to offer some comments to Nisbett’s Op-Ed piece.

    > The first notable public airing of the scientific question came in a 1969 article in The Harvard Educational Review by Arthur Jensen, a psychologist at the University of California, Berkeley. Dr. Jensen maintained that a 15-point difference in I.Q. between blacks and whites was mostly due to a genetic difference between the races that could never be erased. But his argument gave a misleading account of the evidence.

    This is not true. There was nothing misleading, nor incorrect. After 37 years, Jensen’s observations have been clearly established as correct.

    > In fact, the evidence heavily favors the view that race differences in I.Q. are environmental in origin, not genetic.

    False. Anyone who does not understand this may wish to read these:

    Rushton, J.P. and Jensen, A.R. (2005). Thirty Years of Research on Race Differences in Cognitive Ability. Psychology, Public Policy, and Law, Vol. 11, No. 2, 235-294. [easily located on the web]

    > The hereditarians begin with the assertion that 60 percent to 80 percent of variation in I.Q. is genetically determined. However, most estimates of heritability have been based almost exclusively on studies of middle-class groups.

    The B-W IQ gap is larger at the highest level of SES than at the lowest. The above comment is inane. There have been literally hundreds of IQ studies of various racial groups from many different countries and they show the same outcome:

    highest IQ: Ashkenazi Jews
    next highest: East Asians (Mongoloids)
    next highest: Whites of European ancestry
    lower: American Indians and Latinos
    lower still: American and European Blacks
    very low: sub-Saharan Blacks
    lowest: Bushmen and Pygmies

    > For the poor, a group that includes a substantial proportion of minorities, heritability of I.Q. is very low, in the range of 10 percent to 20 percent, according to recent research by Eric Turkheimer at the University of Virginia.

    Turkheimer reached this conclusion by studying children who were age 7 and less. This is long before the shared environmental component vanishes.

    > This means that for the poor, improvements in environment have great potential to bring about increases in I.Q.

    For children only. The shared environmental component equals ZERO for adults.

    > Nearly all the evidence suggesting a genetic basis for the I.Q. differential is indirect.

    Not true. You have to wonder why this person makes such obviously false assertions. Perhaps he thinks his readers are not familiar with the research. A good bit of the evidence is so specific that it cannot be associated with environmental causes by any means.

    See either of the two references I previously gave.

    > There is, for example, the evidence that brain size is correlated with intelligence, and that blacks have smaller brains than whites. But the brain size difference between men and women is substantially greater than that between blacks and whites, yet men and women score the same, on average, on I.Q. tests.

    The brain size difference between the sexes is a virtually perfect predictor of the IQ difference between the sexes. The claim that the sexes have identical IQs is at odds with numerous independent findings, although the subject is still being debated.

    > Likewise, a group of people in a community in Ecuador have a genetic anomaly that produces extremely small head sizes ­ and hence brain sizes. Yet their intelligence is as high as that of their unaffected relatives .

    This person does not understand correlations. Even large numbers of exceptions do not invalidate correlations. One or two points will have little significance, other than to move the correlation coefficient by a tiny amount.

    > About 25 percent of the genes in the American black population are European, meaning that the genes of any individual can range from 100 percent African to mostly European. If European intelligence genes are superior, then blacks who have relatively more European genes ought to have higher I.Q’s than those who have more African genes. But it turns out that skin color and “negroidness” of features ­ both measures of the degree of a black person’s European ancestry ­ are only weakly associated with I.Q. (even though we might well expect a moderately high association due to the social advantages of such features).

    The Black admixture formula (from Lynn) is
    IQ = 80 + (admixture % x 0.2)
    [see Lynn, Richard (2006). Race Differences in Intelligence: An Evolutionary Analysis, Washington Summit Publishers, Georgia.]
    This formula predicts the regional and mean IQs of Blacks in the US and Europe.

    > During World War II, both black and white American soldiers fathered children with German women. Thus some of these children had 100 percent European heritage and some had substantial African heritage. Tested in later childhood, the German children of the white fathers were found to have an average I.Q. of 97, and those of the black fathers had an average of 96.5, a trivial difference.

    This citation is known as the Eyferth study. Environmental proponents always cite it because they have not found any other evidence to cite. But their claim is incredibly lame:

    * The “study” consisted of a very small N. Some citations claim 98 and some 69.
    * Although the children’s IQ was measured, the parents’ IQ was not measured (neither was the rank of the US military father measured). So it is not known whether the children inherited the parental IQ.
    * About 30 percent of US blacks failed pre-induction mental tests for the military, compared with 3 percent of white. So US black soldiers were a more IQ-selected and less-representative sample of their population than were white soldiers.
    * Children were tested prior to the age at which the genotypic aspect of intelligence has become fully manifested.
    * 20% to 25% of the Black fathers were not African Americans but French North Africans (i.e., largely Caucasian or Whites as we have defined the terms here).

    > But it turns out that skin color and negroidness_ of features ­ both measures of the degree of a black person’s European ancestry ­ are only weakly associated with I.Q. (even though we might well expect a moderately high association due to the social advantages of such features).

    There is no presently available study (to the best of my knowledge) of within group IQ variation as a function of skin color. But there is a between group study, as cited in my previously referenced article:

    Templer and Arikawa (2006):

    skin color to winter high temperature r = .85 (p < 0.001)

    IQ to skin color r = – .92 (p The closest thing to direct evidence that the hereditarians have is a study from the 1970s showing that black children who had been adopted by white parents had lower I.Q.’s than those of mixed-race children adopted by white parents. But, as the researchers acknowledged, the study had many flaws; for instance, the black children had been adopted at a substantially later age than the mixed-race children, and later age at adoption is associated with lower I.Q.

    All adopted children reached adult IQs that were equal to their biological peers and which had no correlation with their adoptive families. Transracial studies were not limited to Blacks adopted by Whites but included Asians adopted by Whites. The Blacks ended up with lower IQs than their adoptive families and the Asians ended up with IQs higher than their adoptive families.

    > That environment can markedly influence I.Q. is demonstrated by the so-called Flynn Effect.

    No, it is not.

    > James Flynn, a philosopher and I.Q. researcher in New Zealand, has established that in the Western world as a whole, I.Q. increased markedly from 1947 to 2002.

    Not true. What he found was an increase in raw scores. At the same time scholastic-component raw scores (within the same tests) declined.

    > In the United States alone, it went up by 18 points. Our genes could not have changed enough over such a brief period to account for the shift;

    He got one right.

    > it must have been the result of powerful social factors.

    No. There is absolutely no evidence that any social factors were at work. It appears that most or all of the gains were specific and not g loaded and are apparently due to multiple factors such as nutrition, family size, and environmental conditions that exist in very early life, since the secular rise is observable in toddlers.

    > Most important, we know that interventions at every age from infancy to college can reduce racial gaps in both I.Q. and academic achievement, sometimes by substantial amounts in surprisingly little time.

    The IQ boosts that have been reported were temporary (very short lived) and were believed to be largely the result of teaching to the test. That means that the gains were in s loading and not in g loading. No g loading gains have been demonstrated.

    > This mutability is further evidence that the I.Q. difference has environmental, not genetic, causes.

    No, it is not. Nisbett needs to read (or read again) Rushton, J.P. and Jensen, A.R. (2005). Thirty Years of Research on Race Differences in Cognitive Ability. Psychology, Public Policy, and Law, Vol. 11, No. 2, 235-294.

    • Chuck says:

      Reading that commentary added zero to my understanding. Hey, do you want to help with some analyses? There are a couple of data sets that I need to go through — all public use, all analyzable by free software.

      • Kiwiguy says:

        I didn’t expect it too 🙂 I just thought it was a nice succinct comment on Nisbett’s piece.

        ***Hey, do you want to help with some analyses?***

        Unfortunately, between parental duties and end of year work demands I braely have time to read your site at the moment.

  2. 猛虎 says:

    The Moore study is here.
    I do not understand why it is so widely cited, given that the sample is incredibly small (46) without no information about the IQ of the fathers, and no follow up.

  3. 猛虎 says:

    I agree, something is wrong with that picture (your first link), but it looks different from what I see in the original post.
    See :

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s