Comments Off on
Some ways that statistics can lie to you without ever falsifying their data:
- Reframing a small absolute difference as a large relative difference. For example, it’s technically correct to describe an increase from 0.01% to 0.04% as a 300% increase – but is a +0.03% bump what you’d assume to be the case if you heard the phrase “300% increase”?
- Using one type of average while implying that you’re talking about a different type of average. For example, if a company has nine employees who make $10/hour and one manager who makes $60/hour, you can take the mean average and state with technical truthfulness that the average wage is $15/hour, even though no actual person makes that.
- Reporting accuracy in terms of the absence of false negatives without considering the presence of false positives. A test that returns a positive result 95% of the time when what it’s looking for is present may sound reasonably accurate – but what if I told you that it also returns a positive result 30% of the time when what it’s looking for isn’t present?