What drives audit sample size, and how do statistical and judgmental sampling differ?

Sample size is driven by the desired confidence (inverse to acceptable sampling risk), the tolerable misstatement/deviation rate (smaller tolerance → larger sample), the expected misstatement/deviation (higher expected → larger sample), and population variability; for substantive tests it also relates to materiality and the assurance needed from other procedures. Statistical sampling uses probability-based selection (e.g., monetary-unit sampling, which gives larger items a higher selection chance) and lets you quantify and project the sampling risk and the misstatement mathematically — defensible and objective. Judgmental (non-statistical) sampling uses the auditor's judgement to select and size, often targeting riskier items; it's flexible and efficient but doesn't let you statistically project results across the population. Both must be representative for projection; the choice depends on the population, the need to quantify sampling risk, and efficiency. Either way, you evaluate misstatements found and project to the population, then consider qualitative factors.