MINI LECTURE 15 - Conditional vs. unconditional correlation: twin studies overestimate heredity.

Summary

The speaker critiques the use of heritability equations, particularly in the context of IQ and its predictive power in real-world outcomes. They argue that these equations overestimate the role of genetics due to a lack of understanding of correlation in non-linear environments and the intransitivity of traits. The discussion delves into the flaws of twin studies and the complexities of separating environment from genetics, emphasizing the intellectual shortcomings of those who oversimplify these concepts.

Takeaways

• 🧬 The script discusses the flaws in heritability studies, particularly in the context of intelligence as measured by IQ tests.
• 📊 It criticizes the over-reliance on correlation in non-linear environments, where metrics like IQ may not be informative.
• 🤔 The speaker questions the predictive power of IQ in real-world scenarios, suggesting that high correlation does not necessarily equate to causation.
• 🔍 The script points out that heritability studies often fail to account for environmental factors, which are inseparable from genetic influences.
• 🧭 The concept of 'H Square' is mentioned, which is a measure based on the difference in correlations between identical and fraternal twins, but the speaker implies there may be issues with this approach.
• 🌐 The importance of considering different environments when evaluating heritability is emphasized, as the same trait can have different expressions in different settings.
• 🔄 The script mentions the complexity of extracting unconditional correlations from conditional ones, especially when dealing with different variances across environments.
• 📚 The speaker's upcoming book on entropy is hinted at, suggesting a focus on information-based methods as opposed to traditional correlation metrics.
• 🤖 The script touches on the misuse of statistical methods like regression in heritability studies, indicating a lack of understanding of these tools by some researchers.
• 🌍 The impact of different environmental exposures on the expression of traits like diabetes is used as an example to illustrate the difficulty in separating genetic and environmental influences.
• 📉 The speaker suggests that financial and mathematical finance have more advanced understanding of stochastic correlations, which could be applied to genetic studies.

Q & A

What is the main topic of discussion in the provided transcript?

The main topic is the role of heredity and the flaws in the equations used to determine it, particularly in the context of IQ and its predictive power in the real world.

Why do the equations used to determine heredity tend to overestimate its role compared to the environment?

They overestimate because heredity and environment are not separable, and the equations often fail to account for non-linear environments and the issues with correlation in such contexts.

What is the issue with using IQ as a metric for real-world performance?

The issue is that while IQ might show some correlation with real-world performance, it can be misleading in non-linear environments and does not necessarily predict success in all aspects of life.

What is the problem with relying on correlation in metrics?

The problem is that correlation can become uninformative in non-linear environments, and one should always be suspicious of metrics that rely on it without considering the context.

What is the speaker's stance on the work of Herrnstein and Murray?

The speaker criticizes their work, pointing out that it has flaws, particularly in its assumptions about the predictive power of IQ and its understanding of probability and intransitivity.

What does the speaker suggest about the use of SAT scores as a proxy for IQ?

The speaker suggests that using SAT scores as a proxy for IQ might not be accurate because it does not account for the circularity and potential weaker correlation when applied to real-world performance.

What is the H^2 measure mentioned in the transcript, and what is the issue with it?

H^2 is a measure that calculates the difference between the correlation of identical twins and fraternal twins to estimate heritability. The issue is that it oversimplifies the complexity of genetic and environmental influences.

What is the concept of 'conditional heritability' mentioned in the transcript?

Conditional heritability refers to the heritability observed within specific environments or conditions, which may not represent the heritability across all environments.

Why is it difficult to extract the unconditional heritability from twin studies?

It is difficult because the studies often focus on specific environments or conditions, and the variances in different environments can affect the correlation and heritability estimates.

What is the speaker's view on the use of characteristic functions in understanding heritability across different environments?

The speaker suggests that characteristic functions can help extract the overall correlation from the characteristic function of the overall distribution, which can provide a more accurate understanding of heritability across different environments.

What is the speaker's final critique of the field of psychology regarding the use of IQ tests?

The speaker criticizes the field for its illusion that IQ tests are general and domain-specific, stating that the definition of intelligence required to function changes according to environments, and thus the heritability for specific tests would change as well.