Don't Panic: Probability and Statistics

Probability and statistics is the discipline of making carefully limited claims when the world has declined to provide complete information. Probability starts with a model and asks what data it would produce. Statistics starts with data and asks which models can survive contact with it. This is less mystical than it sounds, although the notation has made a determined effort to look mystical.

The first useful distinction is between an event, a set of outcomes, and a random variable, a number assigned to those outcomes. From there, probability supplies rules for combining uncertainty. Conditional probability asks what changes after you know something. Bayes' theorem updates a prior assessment with evidence. It is the same accounting move behind a medical-test result and a spam classifier, which is a surprisingly broad career for one fraction.

Data arrive as distributions, not as a polite queue of representative facts. A mean describes center, a standard deviation describes spread, and a histogram can reveal a gap or long tail that both numbers miss. The Central Limit Theorem explains why averages often have a nearly normal sampling distribution, even when individual observations do not resemble a bell curve at all.

Inference is where the caution labels become large enough to read. A confidence interval describes a repeated procedure, not the chance that one fixed parameter is hiding inside one particular range. A p-value reports how unusual data this extreme would be if a stated null model held. It does not certify the alternative hypothesis, practical importance, or a causal story. Statistics is good at quantifying uncertainty. It cannot rescue a biased sample, a confounder, or an experiment that was never designed to answer the question.

Regression is similarly useful and frequently given more authority than it requested. It estimates an association between a response and predictors. The residuals tell you when the fitted relationship has left meaningful structure behind. Causation requires randomization or a credible design, not an especially persuasive line through a scatter plot.

Read the Intro for the full path from probability models to inference. Use Slides when the relationships between sampling, distributions, tests, and decisions need a compact map. Keep the Cheatsheet and Practice Reference nearby for formulas and worked patterns. The Exercise asks you to calculate an interval, test a difference, and then write down why the result still has limits. That final sentence is not decoration. It is the part that keeps arithmetic from impersonating judgment.