error log

Please give examples of EIGHT instances of false or misleading information online, from your own online activities (not from class), from the types of false or misleading errors described in EIGHT DIFFERENT WEEKS OF CLASS.  You need not cover every week, you need not offer more than eight examples, but please give EIGHT examples from DIFFERENT weeks.  Please provide the error and some support for your own determination (better research).  This is an informal writing assignment-- it does not need to be written perfectly but it does need to be readable!

    Each student must create a log of errors they find, representing one error from each day’s worth of errors.  For instance, we will discuss direction problems and omitted variable bias on the same day—students’ logs must offer an example of either one of these errors (NOT BOTH) to satisfy the assignment for that day. 

 8 in total 

Here is a non-exhaustive list of issues that have come up over the course of class, for you to consider if you haven't been putting your error log together this whole time.

1. Read beyond the headline
2. Ask for sources
3. Check the quality of the paper—bias, honesty (chart made available)
4. Make sure it’s an actual news site
5. Check the author
6. Double check the actual sources (any official stats should be finadble)
7. Double check key words—arrests crime, crime vs. violent crime
8. Check Snopes, Politifact
9. Double check the date
10. For surveys check the questions and answers (“Do you like this class a) a lot, b) a ton, c) more than any other class” is not a reliable survey that will give reliable information)
11. Remember wikipedia is often edited by the people you’re looking up and by people with a stake in the argument.

• Looking for general fishinessLook for unfair comparisons (apples and oranges—“urban areas” vs. “metropolitan areas”)
• Changes in technology change what you should expect
• Try to find alternate hypotheses that could explain it—maybe my makeup doesn’t give me migraines, maybe it’s the exhaustion
• Contextualize all numbers (how much caffeine is in 99% caffeine free hot chocolate? How does it compare to things that you know keep you awake? How much fat is in whole milk?  How does that compare to how much fat is in 2% milk?)
• Direction issues
• Ommitted variables
• Spurousness
• Data dredging/data fishing/p hacking
• Right censoring (when time has cut off the correct comparison—remember how hip hop guys seem to die young as compared to country music stars?)
• Means, medians, and modes
  1. Can’t find a mean for a categorical variable
  2. Even for continuous variables, outliers (super rich people) mess with the numbers
• Overreliance on P values—the prosecutor’s fallacy (In New Jersey, matching DNA at the scene only means there’s a 1 in 10 chance that you’re the murderer).
• New, not replicated studies
• Studies that can’t be replicated
• Studies where they won’t show you the data
• Small sample size
• Small effect size
• Too many variables (p hacking)
• Too new a design
• Too few people finding those results in a really hot field
• Retractions
• Open access (predatory journals)
• All the bad science
• Inappropriate comparisons (of mice and men)
• Misleading axes
• Lack of a zero
• Percentages that don’t add up to anything
• Lack of description of chart and what is represented