How Statistics and Data Can Misinform

This example illustrates the tendency to overestimate the likelihood of events with greater "availability" in memory - think about familiarity or exposure. This "availability" can be influenced by how recent the memories are, how unusual, how emotionally charged - or simply by how often you encounter particular information.

The tendency to ignore base rate information - percentage without context may present a distorted picture.
2nd paragraph in the link above supplies the number: 257. Think about this number in the context of 320000000 Americans and 3450000 Muslims in the U.S.

Both links (above) are all about spurious correlations. Everybody knows that "correlation does not imply causation", right? But there is more - or rather - less: displaying unrelated data sets together can make you think that changes in one variable are causing changes in the other. Coincidences and lack of relationship do exist.

The tendency to search for and interpret information in a way that fits our existing preconceptions or expectations. This bias makes you pick information selectively. Everybody does this, including researchers...

The example (2nd and 3rd column of the first page) in this classic paper illustrates the framing effect. People will draw different conclusions from the same information depending on how the same information is presented.

The tendency to rely on - or anchor - the very first piece of information we encounter when making decisions.

Both links (above) refer to "Simpson's paradox" - a discovery that averages can be misleading...The relationship between two groups of data might increase, decrease, or even reverse depending whether the groups are separated or combined.

A case study of extrapolation, or rather over-extrapolation.

"To better understand the persistent presence, diversity, and impact of biases, we are compiling a Catalogue of Biases, stemming from original work by David Sackett". From Centre for Evidence-Based Medicine, University of Oxford.

Precision is about exactitude; accuracy - about the degree with which the data is consistent with the true value. No amount of precision will make up for accuracy - if your data sample is not correct, no amount of statistical tinkering will make it accurate.

Both calculators are accompanied by a short yet clear explanations of concepts.

From HBR.

This short guide provides the questions to ask and identifies the pitfalls to avoid to help you get behind news stories that use statistics.

A simple exercise illustrating concepts like sample size or representativeness of data. By Cardiff University.

a/a

Fallacies list with examples. From Texas A&M University Writing Center.

Online since 2001, by Gary N. Curtis - see "About the Author" link on the website.

By John Cook. Start with link #1

The Google Public Dataset Search makes large datasets easy to explore, visualize and communicate.

Zanran helps you to find ‘semi-structured’ data on the web. This is the numerical data that people have presented as graphs and tables and charts. The core technology is patented computer vision algorithms that decide whether an image is numerical. Free registration required.