Summary
The episode works through four questions that students often find difficult in exploratory data analysis: when a point beyond the whiskers of a boxplot is a genuine outlier, why a complete-case file can lose a large share of respondents, what a correlation and its square can say about two variables, and whether the age pattern in the Canadian Social Connection Survey briefing reflects age or the mix of people in each group. It adds worked examples on boxplot cut-offs, compounding missing data and squared correlations, together with a harder gender-mix check of the age difference that uses the counts and means from the briefing file. It draws on a meta-analysis of sleep duration and mortality to show how a correlation can miss a U-shaped relationship, and it compares the survey's age pattern with Statistics Canada's 2021 Canadian Social Survey. The hosts close by discussing whether exploration can inflate false-positive findings, with reference to the garden of forking paths and a 2011 simulation study, and they agree on a practical rule that separates checking the data from finding hypotheses.