Summary
The episode works through four questions that students commonly find difficult in data cleaning and descriptive analysis: whether values beyond the interquartile range fences are errors, what missing at random means and why no test can establish it, how small amounts of missingness compound under complete-case analysis, and what a standardized difference in Table 1 adds to a p-value. Worked examples cover fences for a skewed length-of-stay variable, the share of complete rows and the widening of confidence intervals when several variables have missing values, and standardized differences for two hypothetical studies of very different sizes. Canadian examples include the response rate of the 2011 National Household Survey and the use of tax and benefit records for income in the 2016 Census. The hosts close with a debate about whether health care costs should be summarized with the median or the mean, drawing on Ontario data on high-cost users, and agree that the choice of summary should follow the question being asked.