Probability & Statistics
Understanding Statistics
Develop the skills to interrogate statistical claims you encounter in news headlines, pharmaceutical ads, corporate earnings reports, and political campaigns. These exercises train you to spot the specific techniques -- from axis manipulation to cherry-picked comparisons to survivorship bias -- that make misleading numbers look convincing. You will learn to ask the right questions before accepting any statistical claim at face value.
Context
Why this exercise
Statistics is the discipline of extracting meaning from data while quantifying the uncertainty around the meaning. The basics — mean, median, mode, standard deviation, sample size, confidence intervals — are deceptively simple, and the same numbers can be calculated correctly while supporting wildly different conclusions depending on how they are interpreted. This exercise drills the interpretive skill: distinguishing what the numbers literally say from what people commonly take them to say, and recognizing the standard ways that statistics get used to mislead readers who do not slow down to ask the precise question.
Before you start
Modern statistical inference was developed primarily in the early 20th century by Ronald Fisher (who introduced significance testing, the analysis of variance, and randomization in experimental design), Jerzy Neyman and Egon Pearson (who developed hypothesis testing with explicit type I and type II error rates), and William Sealy Gosset (who developed the t-test under the pseudonym 'Student' while working at the Guinness brewery). The conceptual foundation is that data carry information about underlying populations or processes, but the data are always partial — and quantifying how partial is the work of statistics. A mean alone tells you nothing about variability; a confidence interval tells you the range of population values consistent with the sample; a p-value tells you how surprising the observed data would be under a null hypothesis, but not how likely the null hypothesis is to be true.
Several common interpretive errors recur in everyday encounters with statistics. Confusing mean with median — a single billionaire raises the mean income of a town while leaving the median unchanged — produces misleading impressions of typical conditions. Ignoring sample size leads to overconfidence in small-sample results that fail to replicate. Confusing absolute with relative risk turns a true 'doubles the risk' headline (from 0.001% to 0.002%) into a frightening but practically meaningless claim. Misinterpreting p-values as the probability that the null hypothesis is true (which they are not) underlies a generation of overconfident social-science research findings. And confusing statistical significance with practical significance — a real effect can be statistically reliable and clinically irrelevant — drives over-promotion of marginal results.
The procedural skill is asking the right diagnostic questions before accepting any statistical claim. What is the relevant base rate? What is the comparison group? What is the sample size? Is the reported effect absolute or relative? Is the reported probability conditional or unconditional? Is the apparent pattern statistically robust or could random variation produce it? As you work the scenarios, practice running through this checklist before forming an opinion, and notice when the wrong-answer options describe plausible-sounding misinterpretations that fail one of these diagnostic tests. For broader treatment of how statistics fits into scientific reasoning, see Scientific Thinking.