Thursday, September 23, 2010

The Bell Curve 12 - More Regression Analysis

(It's been a while, so I had to go back through a few older posts, find my place in the books, and then get caught up.)

In the beginning of Part II of the book, "Cognitive Classes and Social Behavior," the authors spend a lot of ink talking about regression analysis[1], in order to explain the conclusions they reach about the effects of intelligence[2] on:

poverty,

unemployment,

single-parent families and illegitimacy,

welfare dependency,

parenting,

crime, and

civic behavior.


It has to be emphasized very strongly from the beginning that:

A, the authors are talking only about non-Latino whites[3],

B, they had a very large data base from which to draw, the NLSY[4], and

C, they divide it into two subpopulations:
1, a high-school diploma, no more and no less, and
2, a college degree, no more and no less.


[1] Regression analysis is a mathematical way of describing the effect of one thing on another thing, in general. If you look at the post from February 6 ("The Bell Curve 4"), with the example of the high-school senior boys lined up by height and weight, you will see again that there is a rough relationship between height and weight -- for boys only, about the same age. It is highly necessary when making correlations, to be very specific about the size of the data set, the constraints imposed upon it, and the conditions under which any conclusions can be accepted as valid.

Another, more technical, way of saying it is:
What is the numerical value of the correlation between the independent variable x and the dependent variable y, if
-1 means a "perfect negative" correlation (every increase in x means a predictable and measurable decrease in y),
0 means no correlation at all, and
+1 means a "perfect positive" correlation (every increase in x means a predictable and measurable increase in y).

[2] That is, the effects of intelligence on various social behaviors, after all other reasonable influences have been accounted for. Other influences could be age, education level, socioeconomic status, parents' socioeconomic status, and so on.

[3] Herrnstein and Murray don't begin to talk about race or ethnicity until part III.

[4] As mentioned before, the NLSY (National Longitudinal Survey of Youth) was begun in 1979 as a nationally representative sample of persons aged 14 to 22 at that time, and who have been "followed ever since." (p. 36, written 1994) It appears it is still going on. See: http://www.bls.gov/nls/

No comments:

Post a Comment