Understanding Studies of Racial DemarcationsOghenovo Obrimah, PhDBlockedUnblockFollowFollowingFeb 3Studies of racial demarcations typically are implemented in context of what are referred to as regression analyses.

Simply put, a regression enables assessments of relations between some variable of interest, say students’ test scores, and variables that define said students, such as race, family income, parents’ professions, parents’ education etc.

Pictorially, with x’s denoting variables that define students, and y denoting students’ test scores, a regression specifies that:Y=f(x_1, x_2, x_3,…, x_n).

This is interpreted as, “y independently is related to each of x_1, x_2, x_3, up until x_n.

”While interpretation as, “y independently is function of (equivalently, is affected by) each of x_1, x_2, x_3, up until x_n” is acceptable, it is conventionally acceptable exaggeration, not always true.

In absence of acceptance of the exaggeration, conditioning of decisions, and this is inclusive of policy decisions that occur within government, on outcomes of regressions is impossible.

Suppose a regression shows ‘students test scores, with test scores as data (y)’ are affected by ‘wealth of parents (x_1), with income or net worth as data’, and that ‘White students score higher than black students, with White=‘Yes’, and Black=‘No’ as data (x_2)’.

Absent any additional information, combined, the outcomes,test scores increase with wealth,andtest scores for White kids are higherdo not imply White parents are richer than Black parents.

Only a study of demographics of the data can reveal White parents are wealthier than Black parents.

Normative Principle One for Interpretation of Regressions:Combined, a relation between y and x_1, and a relation between y and x_2, do not in of themselves imply a relation between x_1 and x_2.

Suppose ‘White parents provide their children with better breakfasts, with identification of different choices of breakfasts that differ in nutritional value as data’, but this is not explicitly explored in context of a regression model.

Suppose also that as already is assumed, ‘White kids score higher on tests than Black kids’.

One could assume that given White kids typically score higher on tests, it can be in part because White parents provide their children with better breakfasts than Black parents.

While this assumption could in fact be true, it cannot be deduced from the relation between test scores and racial demarcations.

A regression makes statements only about data that explicitly is included as variables, does not make any explicit statements about factors that likely are correlated with that data.

To explore this further, suppose White parents are more likely to hire tutors for their kids than Black parents, with ‘hired tutors=‘Yes’ or ‘No’’ as data, and White people are more likely to engage with their children in context of homework assignments, with ‘engage with on homework=‘Yes’ or ‘No’’ as data.

While these factors can be correlated with race, absent inclusion of ‘parents hire tutors for their children, ‘yes’ or ‘no’’ and ‘parents engage with children in context of homework assignments, ‘yes’ or ‘no’’ as variables in the regression, the regression cannot be assumed to render any explicit statements about these factors.

If a researcher seeks to make explicit statements about the additional correlated factors, the factors have to be explicitly included in the regression.

Normative Principle Two for Interpretation of Regressions:Suppose x_2 and x_3 are correlated with x_1, but are not included in a regression of y on x_1 [y=f(x_1)].

While it can be stated that it is believed x_2 and x_3 are sources of relatedness of y and x_1, it cannot be stated that y is related to either of x_2 or x_3.

You may wonder, “why not then include x_2 and x_3 in regression of y on x’s such that the researcher can arrive at the relation between y and x_2, and the relation between y and x_3 alongside the relation between y and x_1?”Pictorially, why not specify:y=f(x_1, x_2, x_3), as opposed to y=f(x_1)?The Problem?Given each of x_2 and x_3 are correlated with y, inclusion of x_1, x_2, and x_3 in the same regression renders any results obtained from the regression less credible.

Under stated conditions, deterioration of credibility of outcomes is a stylized fact of regression analyses.

Among experts, this problem is referred to as problem induced by ‘multicollinearity of x_1, x_2, and x_3’.

If the researcher truly believes x_2 and x_3 matter, he or she would have to explore relations between y and x_2, or y and x_3 separately, or find a way to ‘instrument’ x_2 and x_3, or induce ‘orthogonalization’ of each of x_2 or x_3 from x_1.

But then this gets technical, way over the head of the average reader who is not familiar with regression analyses.

The fact of the matter?.Researchers sometimes have preconceived opinions about things, with outcome they make deductions that contradict the two normative principles for interpretation of regressions that I outline.

The onus is on a reader to be aware of these principles, as such be able to discount any interpretations that are contradictory to outlined principles.

In context of racial demarcations, the right use of regression analyses is for arrival at better understanding of differences between races, for knowledge of marked differences that can be mediated by education or wealth.

The finding for example that improvements to nutrition value of breakfasts improves performance of children in Elementary School has resulted in initiatives targeted at helping children from relatively poor backgrounds have better breakfasts.

Given importance of nutritious breakfasts deteriorate with age of children, benefits of nutritious breakfasts are not available later in life.

If a child does not receive the benefits of a nutritious breakfast while in Elementary School, benefits lost cannot be reclaimed later in life.

A fetus cannot develop in the fourth trimester what it ought to have developed in course of the third trimester.

This is the reason delivery is induced whenever it’s duration inexplicably goes way past the third trimester.

With respect to medical health, the finding that African Americans are more susceptible to diabetes than White folks has led to more of a focus on development of remedies and preventive diets targeted at African Americans.

Structured and used rightly, studies of marked demarcations across races can be a force for good in society.

Given it is fact that the distribution of intelligence is the same across races, studies of racial demarcations that attempt to infer demarcations of intelligence violate data that is factual and contradictory.

The Facts?There are White and Black dropouts from High School.

There are White and Black dropouts from College.

There are White and Black Graduates.

There are White and Black PhD's from top universities in the Sciences, Engineering, Business etc.

There are more White people with Noble prizes in Economics, but there already is one Black man.

It is true factors, such as wealth, income levels, and education levels of parents can affect performance of children in school, but then these factors merely shift distributions of test scores to the left or right, maintain same distribution.

We have then that rightly interpreted, studies of racial demarcations explore factors that shift the same ‘Bell Curve’ distribution of intelligence that obtains within distributions of Whites or Blacks to left or right of each other.

If the distribution of intelligence is the same across races, studies of systemic differences of intelligence across races contradict factual evidence.

Research evidence that contradicts factual evidence never can be reliable.

Research in area of genetics concurs that there does not exist any reliable evidence for presence of systemic genetic differences in respect of native intelligence across races.

If readers of research papers, and researchers make it a point of duty to interpret outcomes of regression analyses with responsibility, without violation of Principles of Interpretation, the probability that discussions of research outcomes are not biased improves for benefit of society.

.