-
Notifications
You must be signed in to change notification settings - Fork 0
/
fisher.html
90 lines (80 loc) · 6.32 KB
/
fisher.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Fisher Doubles Down</title>
<link rel="stylesheet" href="fisher.css">
</head>
<body>
<header>
<h1>Fisher Doubles Down</h1>
<p class="subtitle">When Statistical Tools Support Prejudiced Beliefs</p>
</header>
<section id="introduction">
<p>
In the aftermath of World War II, the UNESCO Commission's publication <i>"The Race Concept"</i> offered a thorough examination of racial classifications and their far-reaching impacts on society. It notably asserted:
</p>
<blockquote>
"Race hatred and conflict thrive on scientifically false ideas and are nourished by ignorance."
</blockquote>
<p>
Despite this, figures like Ronald Fisher remained committed to using statistical methods, such as ANOVA, to reinforce their views on human differences. Fisher believed that genetic variations among populations had a significant influence on intellectual and emotional capacities. His use of statistical techniques gave his subjective views an appearance of scientific validity.
</p>
</section>
<section id="statistical-methods">
<h2>The Role of ANOVA in Reinforcing Bias</h2>
<p>
Fisher's application of ANOVA (Analysis of Variance) was used to "prove" differences across human populations. However, the results of such analyses are only as unbiased as the data and assumptions behind them. When researchers apply subjective weight to certain variables, the analysis becomes skewed, even if the statistical methods are mathematically valid.
</p>
<p>
The F-statistic and p-values are often employed to provide a veneer of objectivity to these analyses. The F-statistic, which measures the ratio of variance among groups to variance within groups, and the p-value, which indicates the likelihood that observed results are due to chance, can seem complex and daunting. Many people can be easily bamboozled by these metrics, interpreting them as definitive proof of significant differences between groups, without considering the underlying biases in how the data was collected or how the variables were weighted.
</p>
<p>
Below are two sets of data for five hypothetical countries. The first dataset introduces bias by favoring 'IQ scores' and 'Moral character score' for Country A and Country C, similar to how researchers might unintentionally (or intentionally) give more weight to certain traits they deem valuable. The second dataset is unbiased, with random scores across all variables, representing a more neutral approach.
</p>
</section>
<section id="biased-dataset">
<h2>Biased Dataset - Emphasizing 'IQ scores' and 'Moral character score'</h2>
<iframe src="biased_data_plot.html" style="width: 100%; height: 500px;"></iframe>
<p>
In the biased dataset, we see that the IQ scores for Country A and Country C dominate the other variables, skewing the analysis in favor of these countries. This shows how statistical methods can be manipulated to support subjective judgments by prioritizing specific variables. In a world where these values are seen as the most important indicators of a country's worth, Country A and Country C would be celebrated, while other countries could be marginalized.
</p>
</section>
<section id="unbiased-dataset">
<h2>Unbiased Dataset - Random Scores</h2>
<iframe src="unbiased_data_plot.html" style="width: 100%; height: 500px;"></iframe>
<p>
The unbiased dataset presents a more balanced view, with all countries performing similarly across a range of variables. No single country or variable dominates, which allows for a more objective interpretation of the data. The p-values in this case are much higher, indicating that any differences between countries are likely due to random chance rather than actual significant differences. This demonstrates how unbiased data produces results that don’t lead to the overemphasis of certain traits over others.
</p>
</section>
<section id="anova-results">
<h2>ANOVA Results Comparison</h2>
<h3>Biased Dataset:</h3>
<ul>
<li><strong>IQ scores:</strong> F-statistic = 45.05, p-value ≈ 2.87e-32</li>
<li><strong>Violent crime per 100 people:</strong> F-statistic = 11.73, p-value ≈ 4.14e-09</li>
<li><strong>Economic output:</strong> F-statistic = 146.65, p-value ≈ 1.30e-82</li>
<li><strong>Moral character score:</strong> F-statistic = 194.60, p-value ≈ 4.17e-100</li>
<li><strong>Reproduction rates:</strong> F-statistic = 63.98, p-value ≈ 1.37e-43</li>
</ul>
<h3>Unbiased Dataset:</h3>
<ul>
<li><strong>IQ scores:</strong> F-statistic = 0.900, p-value = 0.464</li>
<li><strong>Violent crime per 100 people:</strong> F-statistic = 0.917, p-value = 0.454</li>
<li><strong>Economic output:</strong> F-statistic = 0.538, p-value = 0.708</li>
<li><strong>Moral character score:</strong> F-statistic = 0.940, p-value = 0.441</li>
<li><strong>Reproduction rates:</strong> F-statistic = 1.507, p-value = 0.199</li>
</ul>
<p>
The results from the biased dataset show much lower p-values, which might lead someone to believe that the differences between countries are significant. However, these differences arise from biased weighting of certain variables. In contrast, the unbiased dataset has much higher p-values, indicating that the differences observed are likely due to random variation and not significant.
</p>
<p>
These statistical techniques—when presented without context—can easily mislead people into accepting flawed or biased conclusions. By emphasizing certain variables over others, researchers can manipulate F-statistics and p-values to make results seem more significant than they truly are. It's crucial to critically assess the assumptions, data collection methods, and variable weighting behind any statistical analysis.
</p>
</section>
<footer>
<p>© 2024 Colin Geraghty. All rights reserved.</p>
</footer>
</body>
</html>