Study Notes on Correlation and Causation
Overview of Correlation and Causation
The session will primarily focus on correlation, with discussions on how to infer causal claims based on correlations.
There will also be mention of supplementary readings, especially regarding biases in authorship in philosophy and discussion of gender representation.
Gender Representation in Philosophy
Issue raised about male-dominated authorship in discussions of demarcation in philosophy.
Mohammed's input suggests the focus should be on whose voices are heard rather than women choosing not to write.
Recent statistics show women's voices are still not adequately represented in major philosophy journals, particularly in philosophy of science and demarcation.
Noteworthy fields: philosophy of biology is more egalitarian, where the lecturer feels more included.
Ethics: more women’s voices, often regarded as the 'softer' branch of philosophy.
Importance of recognizing biases still present in academic publishing in both philosophy and science.
Demarcation Issues
Demarcation: trying to delineate science from pseudoscience is complex.
A discussion on whether a single criterion alone can suffice to distinguish science from pseudoscience.
Philosophers, especially Popper, emphasized a theoretical basis for demarcation over practical, experiment-based evidence.
The Nature of Theories
Problem is the focus on grand theories like Newton's gravitation and Einstein's relativity, instead of practical scientific methods.
Philosophers often undervalue important empirical details.
Course emphasis will be on scientific practice rather than strictly theoretical discussions.
Correlation vs. Causation
Definitions:
Correlation: two variables change together, which can either be positive (both increase) or negative (one increases, the other decreases).
No correlation indicates that there is no relationship in the data.
Importance of distinguishing correlation from causation, with common pitfalls outlined.
Issues with drawing conclusions from correlations without careful consideration.
Example Analyses
The Pirate and Temperature correlation: Trends indicate rising temperatures and decreasing pirate numbers - an example of a spurious correlation that does not indicate a causal relationship.
The historical context of piracy is crucial to analyzing this correlation.
Cell Phones and Cancer: Misleading claims have emerged from misinterpreting trends in data.
Vaccines and Autism: The lecturer discusses a notable fraud where correlations led to false causal claims about vaccines and autism diagnoses.
Understanding Causal Claims
Correlations can arise from:
A causing B.
B causing A (reverse causality).
Common cause (factor C influences both A and B).
Pure chance or spurious correlations where A and B are unrelated.
The Exercise of Causal Reasoning
Exploring various scenarios where causal relationships are interpreted, including:
Exercise and mood changes: the complex interaction can reflect reverse causality or common causes like stress.
Emphasis on understanding that not all correlations are meaningful or indicative of causation.
The Role of Experimental Evidence
Barrowman’s assertions on experimental science recognizing the need for observation and experimentation to establish causal relationships.
Counterfactual reasoning: exploring 'what-if' scenarios to understand causal pathways.
Bradford Hill Criteria for Causation
Introduced by Bradford Hill in a 1965 paper focused on smoking and lung cancer:
Strength of the association: Measure of correlation.
Consistency: Outcomes in varying situations.
Specificity: Focus on whether one cause leads to one effect.
Temporality: The cause must precede the effect.
Gradient: Dose-response relationship where more cause leads to more effect.
Experimental Evidence: Need for interventions and testing.
Plausibility: Does the causal hypothesis align with known science?
Coherence: Should not conflict with existing knowledge.
Analogy: If a similar cause leads to a similar effect, it adds credence to the claim.
Critiques and Counterarguments
Noted critiques of Bradford Hill's guidelines from various scholars, highlighting challenges in decisively proving causal paths in complex systems.
Importance of critically evaluating both Bradford Hill’s criteria and their critiques.
Concluding Thoughts
Emphasized the significant philosophical implications and challenges in accurately inferring causation from correlation, as outlined in both the main and supplementary readings to be discussed further in future classes.
Next Steps
The next class will broaden discussions on evidence hierarchies and delve into further readings regarding the intricacies of causal relationships in scientific practice, particularly in relation to social issues.