Case 1: Data Manipulation to Support a Narrative

Data can be tailored to reflect a desired perspective, presenting a skewed version of the truth. For instance, in 1973, UC Berkeley faced a lawsuit for gender bias in its admission practices, which revealed a lower acceptance rate for female applicants—44% for males compared to 35% for females. This apparent discrepancy led to accusations of discrimination against women.

However, further analysis of departmental admission rates revealed an unexpected trend: women were actually favored in many departments. The initial lawsuit failed as data illustrated that while the overall acceptance rate was higher for males, this was due to women applying to more competitive fields with lower acceptance rates.

To clarify:

• Claim 1: Men had a higher overall acceptance rate, indicating a bias favoring them.
• Claim 2: Specific departments showed higher acceptance rates for women, suggesting a bias in their favor.

Both claims could be substantiated with data, illustrating Simpson's Paradox, where trends observed in subsets disappear when viewed collectively.

Case 2: Correlation Versus Causation

The adage "correlation does not imply causation" is frequently cited, reminding us that a relationship between two events does not necessarily mean one causes the other. For example, while there is a correlation between long hair and shampoo usage, this does not mean using more shampoo results in longer hair.

Consider another example: the observed correlation between rising ice cream sales and the occurrence of forest fires. This does not imply that buying ice cream causes fires or vice versa. Instead, heat serves as the common factor influencing both phenomena.

It’s easy to find correlations through statistical methods, but it's vital to recognize that a correlation does not equate to causation. There may be a third variable at play, or the correlation could be entirely coincidental.

Here are some surprising correlations identified by analysts:

• Increased margarine consumption correlates with higher divorce rates.
• The age of Miss America relates to murder rates involving steam or hot objects.
• The appearance of Nicolas Cage in films coincides with drowning incidents in pools.

For more peculiar correlations, check out resources that explore spurious relationships.

Case 3: Statistical Misuse in Legal Contexts

The case of Sally Clark serves as a poignant example of how statistics can lead to wrongful convictions. In 1996, after her two seemingly healthy infants died under similar circumstances, Clark was accused of murder. A statistician testified that the probability of two unexplained infant deaths in one household was 1 in 73 million, leading to her conviction.

However, further investigation revealed that the second child had a medical condition, and the deaths were not independent events. After three years in prison, Clark was exonerated, but the tragedy of her case highlights the dangers of misapplying statistical evidence in legal settings.

Statistics are invaluable for understanding patterns and making informed decisions. Nevertheless, they can be manipulated to mislead, emphasizing the need for careful scrutiny when interpreting numerical data.

To delve deeper into this subject, consider exploring the following resources:

• The Misuse of Statistics
• Correlation vs. Causation
• The Sally Clark Case
• Convicted on Statistics
• Simpson's Paradox

The first video titled "This is How Easy It Is to Lie With Statistics" explores how data can be manipulated to mislead audiences.

The second video, "How to Lie with Statistics," delves into various tactics used to distort statistical information.